Oxford University Press, Amen House, London E.C.

4
GLASGOW. NEW YORK TORONTO MELBOURNE WELLINGTON
BOMBAY CALCUTTA MADRAS KARACHI LAHORE DACCA
CAPE TOWN SALISBURY NAIROBI IBADAN ACCRA
KUALA LUMPUR HONG KONG
BY
ANDRE
BEING
THE WAYNFLETE LECTURES
DELIVERED IN THE COLLEGE OF
ST. MARY MAGDALEN, OXFORD
196 1
OXFORD
THE CLARENDON PRESS
1962
FONETISKA INsrrITUTION
Oxford University Press I962
PRINTED IN GREAT BRITAIN
TO
AUSTIN GILL
P E CE
A
TLE is rarely quite as explicit as its author intends
it to be, and a preface is no unnatural place for some
glossing.
The 'functional' label, which appeared in the title of my
1949 booklet Phonology as Functional Phonetics and which I am
using here again, is probably no less forbidding to large
sections of the philological public than the more usual 'struc-
tural' tag. I am quite aware of this, and if I persevere it is
because I am less inclined to flatter the preconceptions of
some of my readers than to stress what, in my opinion, is
conducive to sounder descriptive methods and a better under-
standing of language.
When reading, with full appreciation, the books and essays
ofsome contemporary descriptivists, I cannot help wondering
at times whether, in their striving for scientific detachment
and their laudable efforts to do away with vague idealistic
phraseology, they are not missing what Sapir, hampered
by the pervading psychologism of his age, gropingly
referred to as 'a basic plan, a certain cut to each lan-
guage'. Fighting 'mentalism' should not consist in denying
the existence of well-established facts, but in showing what
palpable realities stand behind loose prescientific phrasing:
it is fair play, for contemporary structuralists, to ridicule a
phrase like 'the spirit of a language' because it is more likely
to evoke some winged supernatural being than a set of
internal relations; but it is bad policy to ignore or neglect
the fact that any utterance or any segment of an utterance
becomes a linguistic object only inasmuch as it has been
identified as belonging to a given language. Physically, a
glottal stop isjust a noise, but in a segment identified as Arabic
it will be perceived as a fully distinctive unit; if the context is
known to be German it will be conceived as a delimiting
V .. LINGUIS1'IC EVOLUTION
I .. REALISM VERSUS FORMALISM
II .. TOWARDS A FUNCTIONAL SYNTAX
INDEXES
39
I
134
161
page I
s E c
IV.. LINGUISTIC VARIETY
III.. LINGUISTIC TYPOLOGY
PREFACE
trick; in a utterance, it will be cough and nothing
else.. What differentiates a language from another language
is not, as still widely believed, another way of combining
ready-made sounds (or letters!) forevoking the same things
or concepts, with specific but marginal deviations called
'accent' and 'idioms', it is a sui generis organization where
particular sound types are given a certain role in shaping
specific signs .. Each of these corresponds to one of the elements
into which experien.ce has to be analysed before it can be
communicated by means of language and is to be classified
according to the particular functions it assumes in linguistic
communication.. Describing a language consists in pointing
out what makes it different from all others.. Now, since speech
organs; perceptive and thinking abilities would seem to be
much the same throughout mankind, what makes it different
is less the substantial nature of the units it operates with than
the way these units function or, in other words, contribute to
communication..
For offering me the opportunity of presenting these views
of mine, first to an audience of distinguished scholars and
advanced students at Oxford, and ultimately to a wider
English-speaking public, my thanks are due to the President
and Scholars of the College of St .. Mary Magdalen who in-
vited me to deliver a series of six Waynflete lectures in the
winter of 196I and to the Clarendon Press for their willingness
to print an expanded version of these lectures.
I should also like to address my thanks to nlY old friends
Eugene and Sylvia Dorfman for reading and improving the
text of the lectures, and to Mr. John Ross for his careful read-
ing of both the manuscript and the proof..
To Austin Gill at whose instigation the invitation was
launched, who made lny stay at Oxford a congenial
experience, and who read my manuscript before it went to
press, this book is dedicated..
A.. M.
ISM RSUS F ................ y ................... IS
T
HE most remarkable achievement of contemporary
linguistics is probably the final assertion of its legiti-
macy as a completely autonomous discipline with its
own object, aims, and methods. Whereas 'philology' had
never severed the ties that linked it to old texts and classical
education, 'linguistics', as a practically new term in English,
labels its contents as free from any dependence or servitude.
On the Continent, where classical scholars were wont to
distinguish between their philological and their linguistic
pursuits, 'linguistics' was, until Saussure and long after the
publication of his Cours, I largely identified with comparative
grammar, and, accordingly, the difference between the former
and the present status of 'linguistics' is the more striking.
Yet this newly won autonomy is frequently lost sight of,
since many linguists are prone to stress, less the unity and
recent self-sufficiency of their discipline, than its multifarious
connexions with other branches of research, old and new,
humanistic or scientific, such as psychology, logic, anthro-
pology, cybernetics, and electronics. This, of course, mainly
results from the fact that, in the scientific world of today, few
linguists are just linguists and nothing else. Just as, in the
past, a linguist was necessarily a philologist and, more often
than not, the student of some literature, he is now likely to
be an anthropological linguist, a mathematical linguist, a
statistical linguist or, of course, still, a philological linguist.
Plain linguists are few and far between. Being one of them,
I may perhaps be excused for concentrating on language as
my sole subject, limiting my excursions outside of that domain
I Now available in English: Course in General Linguistics, trans!' by Wade
Baskin (New York, 1959).
811928 B
2 REALISM VERSUS FORMALISM REALISM VERSUS FOR!v.IALISM
3
to cases where may be to a better understand...
ing of my own field.
It is not by chance that the establishment of general
linguistics, i.e, the study of language for its own sake,
coincided with the proclamation of the view that the study
and analysis of etats de langue should necessarily precede any
other linguistic endeavour. Language had, until that time,
been mainly considered in its evolution, and, accordingly
no one had tried to conceive of it outside the context of
ever-changing human needs that, at all times, jeopardize the
balance of its economy and lead to the restoration of that
balance in a new form. I At one time identified with logic or
reason, language had, by many, been viewed as a product
of collective thinking and was consequently placed at the
intersection of psychology and sociology. That double depen-
dence accounted for its perpetual instability, nay, made of
this instability one of the basic features of language and
thereby hopelessly blurred the boundaries between language,
as we conceive of it today, and what acts upon it from the
outside. Only the strictly synchronic approach recommended
by Saussure could lead to the foundation of an autonomous
science of language through a drastic severance of all its ties
with physical and psychic reality.
Decades of effort toward perfectly static descriptions,
culminating with Hjelmslev's purely formal analysis," have
failed to convince us that they represent an ideal starting-
point for linguistic research. But without Saussure's figment
of the transversal cut," the perfect autonomy of linguistics
might still be a long way off. To that figment we owe the
conception of a linguistic structure perfectly distinct from the
I On language economy, see George K. Zipf, Human Behavior and thePrinciple
of Least Effort (Cambridge, Mass., 1949), pp. 19-22, and Andre Martinet,
Economic deschangements phonetiques (henceforth Economies (Berne, 1955), pp. 94-
97, and Elements de linguistique generale (henceforth Elements) (Paris, 1960), pp.
182-7-
2 See Prolegomena to a Theory of Language, trans!' by Francis J. Whitfield
(Suppl. to I]AL, vol. 19, no. 1) (Baltimore, 1953).
3 Course, pp. 87-88.
factors
at all times, prepare, out of the language of today, the lan-
guage of tomorrow.
No wonder 'structure' has become the rallying sign of
contemporary descriptivists. It points, of course, .to a co-
herence of the components, but, at the same time, to the
aloofness of the whole from all the rest. 'Function', which is
quite fashionable in some quarters, has no great appeal in
linguistics. It is obviously redolent of the uses to which
languages are put; it suggests contacts with the world at
large, those very contacts we have had to disregard in order
to achieve self-sufficiency. Yet, since we are all agreed that
language works as an instrument towards certain goals, we
can hardly deny that the functioning of that instrument
should be one of our major concerns. Actually, all 'struc-
turalists' reckon with the function of linguistic units: setting
apart a feature as 'distinctive' implies that its function
suffices to make it an object of interest and assign, it to a
definite class. But becoming conscious of the paramount
importance of function in linguistics will normally lead to a
greater respect for reality. And what is meant here by 'reality',
is not any physical or semantic trait which happens to be
singled out, but linguistic reality, that which is recognized as
such because it belongs to a given language where it exerts
a definite function. Function supplies the linguist with a
scale of values that will stubbornly resist any attempt on the
part of the theorist to make facts submit to the requirements
of a method. Function may help in bridging the gulf between
the so-called progressive and traditional groups. It may
bring about contacts that will dispel, on the one side, the
belief that only sluggishness and vested interests are pre-
venting one's opponents from discovering the truth, on the
other side, the conviction that the new trends are incorn-
patible with careful observation and respect for facts.
The most deeply rooted objection to 'structuralism' is,
indeed, that concentration on 'structure', whatever that may
4
REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM
5
be, will necessarily lead scholars away from a close scrutiny
of observable facts and make them disregard whatever stands
in the way of their attempts to set up theoretical construc-
tions. In fact, no one among those acquainted with the
linguistic practice of the last decades would deny that there
is some truth here. We all remember so-called descriptions in
which, at every point, facts were culled from the most various
sources, without any regard for consistency, so as to make
them fit the author's preconceptions. Should this be struc-
turalism, no serious scholar would want to be a structuralist.
But, even if we leave aside that sort of irresponsible juggling,
it cannot be denied that any effort towards establishing a
single method for the treatment of all linguistic facts or the
description of any language whatever, will, almost inescap-
ably, result in giving the same status to things which differ,
not only physically-which would be quite in order-, but
also in their role in the economy of the language.
Language reality is far more varied and far less homo-
geneous than many descriptivists would be willing to con-
cede. At many points it gradually merges into other aspects
of reality, which explains why it has taken such a long time
to secure the autonomy of linguistics. For a linguist who is,
above all, intent upon not jeopardizing this painfully achieved
autonomy, the normal reaction to a situation where the
limits between language and non-language seem to be
blurred is to {j:proceed arbitrarily and draw clear-cut distinc-
tions even across uncertain ground. Once his domain is thus
delimited, he may proceed to submit all its parts to one and
the same treatment. Yet he may, all the time, be painfully
aware of the fact that this type of procedure will more or less
distort the picture he will be drawing. From this he may con-
clude that there is no such thing as 'structure' in language
itself, that what is so called is nothing but a frame invented
by the linguist in order to help himto classify the data. In
other words, a structuralist is not one who discovers struc-
I Cf. W. S. Allen, OntheLinguisticStudyofLanguages (Cambridge, 1957), p. 14.
tures, but one who makes them. I This is, ofcourse, an extreme
attitude, but it clarifies the more average position according
to the actual of structure is, at least, not
postulated.
This highly formalistic approach underlies the practice of
a probable majority of contemporary descriptivists, although
it is professed only by a handful of theorists for whom con-
sistency is a fundamental requirement.
The realistic conception of linguistic structure as a feature
oflinguistic reality is apt to be mistaken for the naive assump-
tion that all that is physically present in speech is part of that
reality, irrespective of whether it has a function and what
that function is. Actually, structure can be found in language
only, as it were, as an aspect of its functioning. The varieties
of function establish among data a hierarchy which involves
distinguishing between cores and margins, for whose de-
scription different methods have to be used. Function is the
criterion of linguistic reality. Our duty is to describe that
reality, and it should be no cause of alarm if one of our
operational devices is found to fail us at a certain point.
These devices, such as phonemes, for instance, do correspond
to definite aspects of linguistic reality as shown by the
speakers' comportment, and we value them in so far as they
do, but no further ..
We need not be ashamed of presenting marginalities as
such in our descriptions, because they are the proof of the
latter's truthfulness. Natural scientists, who approach their
own problems far more soberly than do our theorists, arewell
aware of marginalities in those aspects of the universe it is
their duty to investigate and to describe. Zoologists, for
instance, have gone on registering all phenomena as they
entered the field of their observation and have never tried to
do away with the platypus. Why cannot those who study
man's behaviour become reconciled to its entire range?
Our first illustrations will be borrowed from the field of
phonology.
6 ALISlVI VERSUS FORl\,fALISM REALISM VERSUS FORM.A.LISM
7
According to the phoneme theory, every utterance, in
any language, is totally analysable into a succession of dis-
tinctive units; these units are and their
in the language, is strictly determined. Methods have been
devised for the analysis of utterances into phonematic seg-
ments. It is clear, for instance, that the Spanish word mucho
is made up of four successive phonemes: the segment corre-
sponding to the ch group of the spelling is one and the same
phoneme, although it begins as a stop [t] and ends as a
fricative [8], because that fricative never occurs in Castilian
Spanish without the preceding stop, which implies that this
stop, as such, is automatic and not distinctive.. The reasons
for conceiving of the consonantal complex at the beginning
of English chip as a single phoneme are undoubtedly very
cogent, although not quite as obvious as in the case of
Spanish eli. But there are many cases on record where no
agreement has ever been reached as to whether one should
reckon with one phoneme or two successive phonemes:
some would analyse the word ice into two, others, into three
phonemes. If ice presents two phonematic segments, an jaij
phoneme has to be listed among the phonological units of
the language; this will probably entail a similar decision
regarding jauj as the former of the two segments of out. It:
on the contrary, ice is made up of three segments, out follows
suit and English loses two phonemes.. 1
Faced with such a problem, the formalist will reach an
arbitrary decision. But since he wants to be consistent, he
will formulate his verdict so that it can be used as a prece-
dent if some similar case arises. This, however, should not
be construed as if he were looking in the data themselves for
a justification of his comportment.
The realist may also have to decide on one interpretation
or the other, because of some practical necessity, such as the
I Contrast Trubetzkoy's position in Grundziige der Phonologie (Prague, 1939),
pp. 108-9, with the average Bloomfieldian analytic approach as, for example,
in George L. Trager and Henry L. Smith, Jr., An Outline of English Structure
(Norman, 1951), pp. 19-2 2
requirement of a subsequent statistical study. But he will
explain the reason for his choice, and stress for the reader
the factual impossibility of equating, say, the case of jaij
with that of a clear-cut phonematic duality, as in the case of
jklj in clear, and a clear-cut phonematic unity, as in the case
of the first vowel offather.
If the analysis of utterances into a number of successive
phonematic segments raises some questions whose solution is
not quite obvious, the next phonological step, namely the
identification as one phoneme of segments appearing in
different contexts, is so full of epistemological pitfalls that
most structuralists prefer to sidetrack the problem altogether
by declaring that phonetic likeness should be the only guide.
This, of course, sounds beautifully sensible and is, in practice,
highly satisfactory until one comes across cases where a
sound appearing in a given context seems to stand just half-
way between two of the sounds that are attested elsewhere.
Here again, some arbitrary decision is inescapable, unless
one is ready to restate the problem in functional terms: is
the distinction that generally obtains preserved or not in
the position at stake? At this point, the rank and file of
Bloomfieldians will shudder and exorcize the spectre of the
archiphoneme by pronouncing the magic formula: 'once a
phoneme, always a phoneme l'
-If we now come back to reality, we soon discern that while
no one can question the fact that there is between big and
beg a difference, lying somewhere in the middle of the two
utterances, which suffices to distinguish them, it is far more
risky to state that the two vowels of kitchen are the same. For
a satisfactory functioning of the language, at any rate, it is
essential that the vowel of big and that of beg should be kept
distinct, but it is immaterial whether the two vowels of
kitchen should be the same, or not the same at alL In theory,
there is no reason for the system of word initial consonants
to present any point of similarity with that of intervocalic or
word-final consonants.. We could easily imagine a language
8 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM
9
where all word initials would be articulated in the front part
of the mouth, and all the final consonants at the back and in
the throat, so that it would be meaningless to try to pair
every unit of one system with one of the other. What prevents
this is, of course, economy: people who make distinctive use
of labial stop articulation in one position would be stubborn
monsters if they resisted, throughout millennia, its extension
to other positions..
Where economy requires that systems of distinctive units
should differ from one position to another, systems are differ-
ent, as is shown by the coexistence ofvocalic and consonantal
systems.
It is quite essential to try to understand why we all insist
on calling the [p] of pat and that of tap one and the same
phoneme. We do so for no cogent theoretical and abstract
reason, but only because both appear to be the product of the
same muscular habit with just the necessary adaptations
imposed by the respective contexts. Both are prevented from
drifting by the same inhibitions determined by the necessity
of keeping them distinct from other articulatory habits, the
same ones for the [p] ofpat and that of tap. If the competing
articulatory habits were not the same for initial [p] and
final [p], it: for instance, the language under consideration
were not English, where we have tab alongside tap, but one,
like German or Russian, where no such pair could be found,
some of the drift-preventing inhibitions existing initially
would be found missing finally. Strange as it may sound,
true realists, in such a case, would object to the complete
identification of initial [p] and final [p] even if these could
be shown to be physically identical. What is decisive, in
language, is achieving communication, and this is secured if,
at every point in the utterance, the unit chosen is kept distinct
from the ones that could have been used, in the very same
context, in order to make a different message. But how close
this unit is to one actually used elsewhere is of secondary
importance. It is quite obvious that we are trained, from
infancy onward, to detect in the speech of others what corre....
sponds to distinct choices on their part. Learning to speak is
learning to make the choices current in one's community.'
Phonemes are such choices. They do play their distinctive
role precisely because they are specific choices and recognized
as SUCI1.
It is at the same time normal that there should be different
systems for vowels and for consonants and that the various
consonantal or vocalic systems of a given language should
tend to present similar distinctions, but no one should try to
spirit away any discrepancy between the distinctive pattern
in one position and that in another. Yet formalists have
rationalized some of their arbitrary decisions into a principle
of simplicity, according to which the best language descrip-
tion is the one which presents the fewest possible units. When
they come across a situation where one type of distinction is
found in position A and another type in position B, they never
stop before they have convinced themselves, if not ofhers, that
the two types are nothing but avatars of the same distinction.
The case of French [e] affords a nice illustration offormal-
istic and realistic practice in a case of that sort: Parisian
speakers of my generation are apt to distinguish between un
metre, 'a metre', and un maitre, 'a master' ;2 the distinction lies
in the vowels: both are normally said to be [], with or
without minor tamber variations, the difference being one
of duration or whatever gives that impression. In word-
final checked syllables, such as metre, maitre, the only other
front retracted vowel units are li/ as in mitre, 'mitre', and lal
as in quatre, 'four', which gives a partial system that could be
represented as follows:
1/ (m)etre
Iii (m)itre lei (m)aitre [e] (qu)atre
I Cf. Elements, pp. 31-32
2 Cf. Andre Martinet, La Prononciation du francais contemporain (henceforth
Prononciation) (Paris, 1945), pp. 126-9; the distinction is being abandoned, as
shown in Ruth Reichstein's 'Etude des variations sociales et geographiques
des faits linguistiques', Word, xvi (1960), p. 61.
10 REALISM VERSUS FORMALISM REALISl\1 R.SDS FORMALISM II
III absolute final position, the same French usage distin-
guishes four vocalic qualities, Iii as in riz; lei as in re, lEI
as in rais, lal as in rat. the partial system looks as
follows:
Iii riz lei re 181 rais lal rat
We have thus four distinct units in both positions, and the
principle of simplicity will convince formalists that the final
lei of re must, by hook or by crook, be identified with either
one of the two IE/,s of metre and maitre. Pure and conscious
formalism will favour an arbitrary solution. Mitigated for-
malism will try to buttress its decision by reference to minor
tamber variations. After long hesitations, Roman Jakobson
decided in favour of equating the vowel of re with the short
lEI of metre, whereas the long vowel of maitre was identified
with the lEI of rais whose duration, in the dialect being
discussed, is the normal short one of word-final vowels.
1
Should instrumental observation reveal, in many speakers,
a slightly closer vowel in metre than in maitre, this would
certainly not point to any tendency to make it conform with
the close vowel of re, because if such a tendency existed it is
hard to see why it should not 'be more successful.
Instead of looking from one local system or inventory to
another, we shall, more realistically, consider what happens
in a definite position, namely before word-final -tre, where
vocalic phonemes have, of course, to be kept distinct from
one another if linguistic communication is to be secured.
Next to our two IE/,s, in the system, we find two [a] phonemes,
those of quaire and pdtre; the former is often said to be front
and the latter back, and here lie the essentials of the dis-
tinction in strictly local Parisian pronunciation where the
vowel of pdtre might, in the 1920'S, have sounded like the
I The present writer was used as a native informant, and his testimony was
solicited first in favour of the identification of the vowel of re with that of
maitre, then in support of the grouping which was eventually retained in
R. Jakobson and John Lotz, 'Notes on the French Phonemic Pattern', TtTord,
v (1949), p. 154
-aw of English law.
1
In another of pronunciation, which
has greater prestige, the tamber difference is much smaller,
and it is eked out a difference in length, so that the vowel
of quatre can be said to be short and front, that of pdtre long
and slightly back. If we 110W consider our vowels in the
frame of the articulatory quadrilateral, it should be clear
that the two short vowels, that of metre and that of quatre, are
about as distant as the two long vowels, that of maitre and
that ofpatre, the former pair lying a little higher and more to
the front than the latter, as shown on this diagram:
8
\8"
a a
This is the type of fact on which linguistic realists are likely
to capitalize. In the present case, even in a purely static
presentation of the vocalic structure of French, their aware-
ness of the importance of the relationships among the units
which really form a system, i.e, among which speakers have
to choose at each point, if communication is to be secured,
will prevent them from ascribing too much importance to
the relationships among units from one system to another
and indulging in largely arbitrary speculations as to how
maximal descriptive simplicity is to be achieved.
The case of what is called the French 'mute e'2 gives a
more dramatic illustration of the contrast between what I
am tempted to call formalistic totalitarianism and realistic
discrimination. English spelling is full of silent e's, as in case,
mute, and give. Since they never correspond to any audible
articulation, they do not raise any phonological problems.
In French, some e's are found in the spelling where, accord-
ing to context or style, either nothing is heard, or a more or
less centralized vowel is sounded, which we will henceforth
designate as [d]. It is important to notice that such an alterna-
tion of zero and [d] is never found initially in an utterance or
I Cf. A. Martinet, 'C'est jeuli Ie Mareuc', Romance Philology, xi (1958),
p. 352 . 2 Cf. Prononciation, pp. 37-70 .
12 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM
a word, but always after a consonant.. In normal everyday
unsophisticated spoken Parisian French, the alternation of
zero and [d] is in ninety-nine cases out of a hundred auto-
matically regulated by the context, and this irrespective of
whether some e appears in the spelling or not.. Phoneticians
usually say that the appearance of the [g] vowel is regulated
by the 'law of three consonants', I which prescribes that [g]
should be pronounced where it will prevent clusters of more
than two consonants, thus, for instance, in arquebuse and
contre-maitre, where it finds its expression in the spelling, and
in ours blanc [ursobla], and arc-boutant [arkobuta] where it
does not.. This amounts to saying that 'mute e' is no phoneme
since its appearance is automatically determined by the con-
text, and that for every consonant phoneme of French we
should reckon with two allophones: a purely consonantal one,
the one we find for jdj in dans, and one, partially vocalized,
which we encounter in devant ([dgva], phonemically jdvaj).
All this is, of course, a greatly simplified picture of reality:
there actually are cases in which the pronouncing or not
of 'mute e' may make a difference in the interpretation of
what is said.
First of all, French children in the course of their slow and
protracted learning ofa difficult language, eventually manage
to distinguish between two different treatments of words
beginning with a vowel, the one which prevails in the
enormous majority of cases and which children tend to make
universal, and the other which is reserved to words which
used to be pronounced with initial [h]. Hence l'eau [10], but
le haut [loo] , l' etre [lstr], un etre [&ntr], but le httre [ldtr], un
hare [&trJ. As is shown by some of these examples, the
phonic difference often lies in the presence or absence of the
so-called 'mute e', This implies that [g] cannot be equated
with zero before a word-initial vowel.
The name of the ram, in the medieval epic of Reynard the
Fox, was Belin, and this survives as a surname, but in two
I Maurice Grammont, La Prononciationfrancaise, 7th ed. (Paris, 1930), p. 105.
different forms, Blin and Belin which are phonemically quite
distinct: some people are called Blin and others Belin, and
confusions are not likely. With ordinary nouns, which are
norrnally preceded by an article, the situation is different,
and I have heard children wonder why a small fish (l'ablette)
had the same name as the weasel (la belette) ..
At the age of six, children go to school and begin to learn
fables and other pieces of verse. Now, French metrics are
based upon the language as it was pronounced at the
beginning of the sixteenth century when all 'mute e's' were
sounded or, at least, made a difference, unless they were
elided. A line like
Et Je sais que de moi tu midis l' an passe
which would amount to ten syllables in normal speech
([E3-sEk-dg-mwa-ty-me-di-la-pa-se]) is recited with the regu-
lar twelve syllables of the alexandrine. This contact with
classical poetry is, with the average child, too incidental for
him to learn where to put a 'mute e' and where not to put it.
What he remembers is that one has a right to insert a 'mute e'
after any consonant if metric necessities require extra syllables..
And this is indeed what happens when children sing their
own words to a tune; as, for instance, On t' emmene ce soire ..
instead of the traditional Va chez la ooisine. .. .. . In popular
poetry, the 'mute e's' of the spelling are sounded or not,
according to the occasional needs of the author. A protracted
contact with classical verse will, however, finally condition
people in that respect. Apart from the initial consonant, I
pronounce soir and poire alike; but I would no longer accept
them as rhymes.
The above is a realistic sketch of a most intricate problem.
It is realistic because it indicates certain aspects of the lin-
guistic comportment of French speakers without trying to
account for them all within one and the same formula.
A totalitarian approach would consist in stating that since
there are cases when [g] instead of zero makes a difference,
REALISM VERSUS FORMALISM
l'dl is a and be marked as It IS
actually or potentially there, 'Potentially', of course, is meant
to take care of traditional poetic diction. This type of descrip-
tion, if it deserves to be called a description, has definite
practical advantages: it gives the language a monolithic
appearance which is thought, in many quarters, to make the
approach to French as a second language easier; it meets
wi th the approval ofthe mass ofeducated Frenchmen who are
inclined to identify culture with the written form of their
classical heritage; but its main advantage in the eyes of the
formalist consists in the vistas it opens toward a further
simplification of the pattern: it is clear, for instance, that if
we can analyse cane, 'duck', which is pronounced [kan] , as
/kang/, nothing but phonetic reality (but who cares about
that?) can prevent us from analysing quand, which is pro-
nounced [ka] , as /kan/, and in general analysing nasal
vowels into oral vowel plus nasal consonant. We thus get rid
of the four distinctive units we would have otherwise to posit
in vent, oin, long, and un. It is true that a word like canton will
thereby require six successive phonemes (/kanton/) instead
offour (/kat5/). I But it would seem that formalistic simplicity
applies exclusively to the number of units in the system, not
to that in the utterance.
The main disadvantage of this type ofinterpretation is that
it gives a completely distorted picture of present-day natives'
linguistic comportment. Toward the end of the fifteenth
century, French was a language in which, with very few
exceptions, all syllables ended in a vowel, very much like the
Old Macedonian of Cyril and Method. z Consonantal groups
were unpronounceable unless they existed word-initially: in
a word like absent, the b must have been mute in normal
speech. On the other hand, [g] was the only vowel found in
totally unstressed position, and consequently it could be
dispensed with without any resulting confusions. Through
I cr. Knud Togeby, Structure immanentede la languefrancaise (Travaux du Cercle
Linguistique de Copenhague 6), 1951, p. 58. 2 cr. Economic, pp. 326-8, 34-9-52.
REALISM VERSUS FORMALISM
this gradual elimination of [d] wherever the result was not
too heavy a cluster, biconsonantal groups of all kinds became
pronounceable: gibeciere, 'satchel', formerly a four-syllable
word, was reduced to two with the group /bs/ in between.
From that time on, speakers could yield to the suggestion
of the spelling and sound the -b- in absent. Therefore, the
/-bs-/ cluster in absent is not, in ordinary speech, different
from the same group in gibeciere.
From an educational standpoint, it may suffice to state
that no one can hope to understand spoken French before,
consciously or unconsciously, he has learned to disregard
'mute e' as a distinctive segment except in the very specific
situations which have been noted. It is not our purpose, here,
to examine how this could be achieved.
A last illustration will also be borrowed from French, but
this time I shall consider the case of a grammatical category,
namely gender. Here is a difficult problem which we can
only hope to clarify if we refuse to let ourselves be .rushed
into accepting from the start a totalitarian approach.
My choice of French is largely determined by the fact that,
since there are only two genders in that language, it will be
easier to conceive of the gender system in its totality, if in-
deed we are entitled to speak of it as a system. It goes without
saying that the conclusion at which we shall arrive regarding
French does not necessarily apply to other languages, even
those as closely akin as Italian or Spanish.
French distinguishes between a 'masculine gender' and a
'feminine gender'. The formal behaviour of nouns belonging
to the masculine gender is simpler. If all nouns behaved
like them, there would, of course, be no gender problem,
quite exactly no gender. We may therefore assume that the
feminine gender is the key to gender. We might have done the
reverse and considered normal the behaviour of feminine
gender nouns: then, the masculine gender would have been
the key to the gender problem. But OUf actual choice is
operationally preferable.
First, it is essential to distinguish feminine gender from
reference to female sex: when English speakers say she,
instead of he or it, it is because they intend to refer to a
female being or, exceptionally, to some machine they choose to
consider in the same relation as a female. The language offers
three possibilities. The speaker is determined to choose one
or another by non-linguistic factors. On the contrary, when
French speakers say elle instead of il, they do so because they
cannot do otherwise. Apart from highly exceptional cases
which we shall see later, there is, in such a case, no trace of
a choice on the part of one who pronounces elle. The speaker
is forced to use this or that pronoun because he is referring
to a linguistic unit that requires him to say elle instead of il.
as I believe we should, we refuse to identify as a linguistic
unit a segment that does not correspond to a new choice ofthe
speaker, we must declare that gender in French is no mor...
pheme, The situation is perfectly clear when, leaving aside
the personal pronoun, we concentrate on the marks of
gender in the nominal group. When I say la grande montagne
blanche, I choose to use the definite article and not the
indefinite une; I choose to qualify the mountain as great; I
choose to speak of a mountain and not, for example, of a
curtain; I choose to qualify it as white. But I never choose
feminine instead of masculine. This does not mean that I
refuse to characterize the mountain as a female: I certainly
do. It means that I am not given a chance to choose a gender,
because as soon as I say, or foresee I am going to say, mon...
tagne, I cannot avoid giving the definite article and the
accompanying adjectives the fuller, so-called 'feminine',
form.. My choice of montagne implies a number of formal
accidents which tradition labels as feminine forms. In Saus-
surian terms, we may say that the signifiant of the feminine
noun montagne is not limited to that form, but that it is dis...
continuous and emerges at other points of the utterance. In
our particular instance, it shows as the / .... e] of la, the
I. dl of grande, the I JI of blanche.
16 REALISM VERSUS FORMALISM
REALISM VERSUS FORMALISM 17
Gender is quite frequently lumped together with number,
as two frequent nominal categories. But this should not
induce us to put them on the same level of analysis.. Number,
the plural number, for instance, is, in French, a morpheme:
if instead of la montagne Ilamotan/, I say les montagnes Ilemo...
tanI, I choose to say I. . . e " .. .1instead of I. .. .. a .... / because
I want to stress that several mountains are involved, not one.
Nothing, in the context, compelled me to say les instead of
lao The plural was a new choice.
It might be argued that French adjectives are frequently
used by themselves, both the masculine and the feminine
forms, with the power and function of a noun, and that, in
such a case, the marks of the feminine stand by themselves
without the support of the feminine noun. But, of course, the
adjective, here, is in the feminine because it refers to a sub-
stantive which is understood, and could be furnished on
request. In most cases of that kind, we would find the sub ...
stantive to be actually present if we cared to include the
whole conversation in the context, so that here again the
feminine would not correspond to a new choice.
All this might induce us to state that the class of feminine
substantives in French is nothing but a group of nouns which
share certain formal characteristics. Theoretically, the case
would not be different from, say, that of the English sub-
stantives ending in -ow, such as shadow, widow, morrow,
which have nothing in common except that final ... ow. Of
course, the signifiants of French feminine are discontinuous
and their elements are dependent on the choice of accom-
panying adjectives, whereas the -ow of shadow can never be
separated from shad... and is quite independent of the sur...
roundings. But once these purely external features are set
aside, the contents of the feminine gender boil down to zero,
French feminines having hardly more in common than
English ... ow substantives have.
Other circumstances have to be taken into consideration,
however, before we may venture to draw our conclusions:
811928 C
18 REALISM VERSUS FORJVfALISM
REALISM VERSUS FORMALISM :::9
it must be kept in mind that when a formally ferninine noun
is applied to some animate being, that being is normally a
female. There are, of course, a good many exceptions:
baleine 'whale' is a feminine, which does not imply that the
French believe whales are exclusively females. Still, the
feminine gender of baleine cannot be equated with that of
chaise or table. When referring in English to a chair or a table,
I, a native speaker of French, am not tempted to use she. But,
in reference to a whale, she would naturally be on the tip of
my tongue. For the French, the unicorn is a female for the
purely formal reason that unicorne was interpreted as une
icorne before becoming une licorne. Consequently, it is difficult
for me to get reconciled to hearing, as I recently did in a
film a unicorn referred to as he. There is at least some hint of ,
sex in the feminine gender when used in reference to ,what
has a sex.
Furthermore, we should not disregard the fact that the
feminine form of many adjectives differs from the corre-
sponding masculine in exactly the same way as the designa-
tion of the femalediffers from that of the male: formally,
belle is to beau what chamelle 'she-camel' is to chameau. Jocu-
larly, one could and does derive from bourreau, 'hangman',
a word bourelle for a 'she-hangman'-or should I say 'a
hangwoman'? We really cannot argue that we have, here, a
pure and simple case of homonymy, because both uses of the
-eaul-elle alternations actually fade into each other as shown,
for instance, by the substantival uses of beau and belle.
Last but not least, elle is sometimes used in reference to
females who, in the same context, have been designated by
means of a masculine noun. A female physician is normally
referred to as un docteur, but this never entails the use of the
masculinepersonal pronoun il. In a sentence like Le docteur
est arrive; elle est dans le salon, there is no doubt that the pro-
noun refers to the sex of the person and not to the gender of
its nominal designation. However, it is worth noting that,
in the case of male beings designated by means of feminine
nouns, the gralnmatical concord is : la sentinelle
elle. . . . The use of il in such a case would imply that the
specific reference to the soldier as a sentinel had been for-
gotten. It would seem that the lack of gender agreement in
the case of le docteur ... elle ... is a recent development and
can still be considered a strictly marginal affair0 But it points
to the fact that the use of elle instead of il may be governed
by non-linguistic and this is a first step on the road
followed by she in the course of the history of the English
language.
A totalitarian description of the French language would
capitalize on this still sporadic phenomenon, stress the formal
identities in the expression of sex and gender, and finally
emerge with all the features of the traditional, highly
idealistic, presentation of grammatical gender where it was
assumed that the exhaustive distribution of all concepts into
two classes of masculines and feminines entailed the anima-
tion and sexualization of all conceivable objects.
A discriminating and functional approach will, of course,
stress the actual lack of function in a large majority of cases,
namely when the substantive, whose gender entails some
formal modification of articles and adjectives, is present
among them. It will take into consideration the somewhat
'pronominal' value gender marks will assume when the
substantive is understood or remote in the utterance, and it
will take care of occasional lacks of gender agreement. But no
formulation shall be striven for, which, in an effort to cover at
once all known facts, would infallibly blur and even conceal
the most specific, enduring, and characteristic features of
gender.
It might be tempting, at this point, to indulge in a simile
and say that a formalistic description is to the language what
a geographic projection is to the actual shape of the con-
tinents, in the sense that the centre of the nlap is fairly
accurate, but the margins are distorted, sometimes beyond
20 REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM 21
recognition. But this would give too flattering a picture of
the achievements of formalistic linguistics. What really hap-
pens, in too many cases, is that the inclusion of margins warps
the whole description in their favour at the expense of what
is, linguistically, really vital, as when the basic non-distinc-
tiveness of 'mute e' in French is never even mentioned
because all centres around the exceptional cases where it
assumes a distinctive function.
The danger inherent in the attempts to squeeze all facts,
central or marginal, into the same pattern exists on all
planes of linguistic description. It lurks, from the start, in
attempts to define the object of our science. In theory at
least, the first question a linguist should ask and try to answer
is: What is language? We may, in true Saussurian fashion,
come to the provisional conclusion that language is a system
ofsigns. But, ofcourse, the question arises whether all systems
of signs are languages. Offhand, honest and sedate linguists,
who are trained to operate on such languages as Latin,
Russian, Chinese, but hardly on traffic signals, are inclined
to give a negative answer; but faced with the complexities
of human languages, with central cores and marginalities
that they dare not identify as such, they finally yield to the
pressure of 'annexationists', those who want to put the lan-
guage label on as big a chunk of semiology as they can grasp.
This leaves us without a definition of what it is our duty and
our aim to investigate, namely human language properly so
called in its different forms, the languages as actually spoken
by men.
Ifwe want to know what language is, we should not try to
list all the features we may have come across when studying
the most diverse languages, and draft a definition which will
somehow vindicate them all. We should rather try to deter-
mine what all the languages we know, all the communica-
tive instruments we want to call 'languages', really have in
common, so that we would not be willing to call 'language'
some semiotic system which did not present that minimum.
Languages serve many purposes. They certainly help us to
think. They give us an outlet for our pent-up feelings. We
use them as artistic mediums. But they are first and foremost
used for communication, i.e, transmission of experience from
one person to another. Communication is, of course, in-
volved in the artistic uses we make of language, and what
is not communication there belongs to expression, a phrase
which, in technical parlance, we should reserve to self-
centred linguistic activity which does not aim at transferring
information from speaker to hearer, but to give the former
relief from internal pressures and tensions of all sorts. Soli-
loquy, which is pure expression, is normally frowned upon as
unsocial behaviour, and those who want to 'express' them-
selves, will have to cheat and procure a victim with whom
they can sham communication. This explains why linguistic
evolution is entirely determined by the communicative needs
of man: soliloquy, ifit were not aping communication, would
soon result in the annihilation of language.
It may not be out of place to mention here that when I
speak of experience as being transmitted by means of lan-
guage, I give these terms their broadest possible meaning:
experiencing a wish or a craving is part of experience, and
its transmission may assume various forms: plain statement,
request, or command.
Experience as such, prior to all attempts to transmit it to
others, is not couched in words, except, of course, if it is
gained through linguistic communication. A very immediate
type of experience, such as pain, is a good starting-point for
understanding how and at what point language enters the
picture. Vocal reaction to pain may be purely reflex: a
groan. The groan may also be willed and meant as com-
munication, but it is no linguistic communication: cats do
communicate with their meows, yet we do not want to
include meows in language. Language begins when the
homogeneous unanalysed feeling is interpreted into a suc-
cession of definite vocal stretches, each of which can be used
22 REALISM VERSUS FORMALISM
REALIS1v1 VERSUS FORMALISM
in the transmission of totally different of experience,
but which, when grouped and ordered as we hear them,
convey fairly specific information about what the speaker
feels. Should I say, for instance, I have a headache, I would use
five vocal stretches, namely I, have, a, head, and ache, each of
which can be found in totally different contexts for the
conveying of entirely different things; they are thus most
unspecific, but, when grouped, they reach a fairly high
degree of specificity.
It is not too clear what is meant when laymen describe
human languages as 'articulate'. It is likely that people who
use this phrase just repeat what they have heard without
having ever thought of the implication of that term. Yet it
describes perfectly what characterizes human language, less
in contradistinction to various forms of animal cornmunica-
tion than ill contrast with human experience before it has
been analysed with a view to linguistic communication.
What characterizes linguistic communication and opposes it
to prelinguistic groans is precisely this analysis into a number
of units which, because of their vocal nature, are to be pre-
sented successively in a linear fashion. These are the units
which many contemporary linguists call 'morphemes'. But in
view of the fact that a number of others use 'morpheme' for
different purposes, I prefer calling them 'monernes'. Monemes
are the smallest segments of speech that have some meaning
attached to them. According to Saussurian terminology, they
are minimal 'signs', with two faces: signijiant and signifie.
The way experience is analysed differs from one lan-
guage to another. The set of habits we call a language
suggests the breaking up of experience a number of
elements for which the language in question has equivalents:
a language may use, for headache, a specific moneme, some-
thing like migraine, instead of two. Where the English say
lift, others would speak of a hoisting machine. The way people
proclaim their ignorance is very differently articulated in
English I don't know, in French Je ne sais pas, in German ich
weiss es nicht. differences in articulation show not only in
the way monemes are combined into utterances, but also in
the range of people have at every point : where an
English speaker may choose among blue, green, and grey for
conveying his experience, a Welshman will have to be con-
tent with a single colour designation glas. All this points to
a fundamental feature of human language: its variation from
one community to another and its variation through time.
Throughout the world, cats say meow because this results
from voice accompanied by a lazy opening and closure of
the jaw. Language varies because it suits the varying needs
of man. It follows that any feature of speech that is auto-
matically found in all communities must be considered non-
linguistic or, at best, marginally so. It is not a matter for the
linguist to deal with, but for the psychologist, the physio....
logist, or both, since these scholars study man in general,
assumed to be the same throughout mankind. Our aim, once
we have agreed on the features we want to find in an object
before we list it among languages, is to describe languages,
i.e. to indicate what makes a given language different from
all others, whether these are actually described, known to
exist, or just conceivable. The necessity of taking into con-
sideration languages unknown to us compels us to reckon
with all the possibilities that are not explicitly or implicitly
ruled out by our definition of language.
The articulation of experience into successive units is only
one of the two features which we want to include in our
definition of language. We have to reckon in language with a
second articulation, that ofthe phonic aspect ofevery moneme
into a succession of distinctive units, the phonemes. Every
one of the five units of our former example is formally made
up of one, two, or three sounds or phonic complexes, to
which, as such, no meaning is attached, but whose choice and
order fully characterize the moneme whose manifestation
they are: head, for instance, is made up of the three phonemes
[u], lei, and Idl in this order.
REALISM VERSUS FORMALISM
Linguists of the old school, Saussurians be
tempted to forget or disregard the second articulation. They
would thereby miss some fundamental conditionings of
human communication. The obvious advantage ofthe second
articulation is economy. The first articulation was economi-
cal in the sense that with a few thousands of fairly unspecific
monemes, it was possible to shape an infinity of different
communications. In the same way, the second articulation is
economical, since the judicious combination of a few dozen
phonemes enables man to keep distinct all the monemes he
needs. In view of the great variety and richness of human
communication, the double articulation was bound to be a
feature of human language: let us try to imagine how we
would fare if we, both as speakers and hearers, had to dis-
tinguish the thousands of homogeneous grunts which we
would need for everyone of our monemes if the second
articulation were unknown. It is clear that the lexical ex-
pansion made necessary by the progress of mankind would
have been unthinkable without the tremendous economy
entailed by the breaking up of signifiants into phonemes. But
there is more to it than economy" Ifthe form ofevery moneme
was an unanalysable grunt, there would be a complete
solidarity between sense and vocal form. Meaning would
exert a direct influence on form, and form on meaning, the
result being that at every instant, every speaker would be
tempted to adapt his pronunciation to the particular shades
of meaning he would want to convey to his audience. As a
final result, both form and meaning would be in a permanent
state of wavering, and this would prevent the establishment
of discrete meaningful units, what the monemes of our
languages actually are thanks to their well-defined and
stable forms" The articulation of signifiants into a succession
of phonemes practically excludes the meaning of a given
word from exerting any influence on its specific form. There
exists a solidarity among all the performances of the same
phoneme which tends to preserve its identity whatever the
REALISM VERSUS FORMALISM
sense of the word may be. The phonic environments may
result in distortions or warpings away from the performance
of the phoneme in isolation. But the semantic context is
normally powerless. All this is, of course, nothing but, a
synchronic version of the regularity of phonetic changes:
only in very particular cases, which should be especially
accounted for, may meaning affect phonetic evolution.
This phonematic solidarity can only be explained if we con-
ceive of the phoneme as an articulatory habit, something we
should always keep in mind, even if we choose to disregard
it temporarily when concentrating on descriptive methodo....
logy.
A problem which always arises when language is to be
defined is whether the vocal nature of speech should or
should not be included in the definition. On the one hand,
it is a fact that the languages linguists deal with are or were
primarily spoken, even when we can only approach them
indirectly through written texts, and that, for a deeper
understanding of those written texts, we should always try
to imagine the spoken medium upon which the written
literary form was based" On the other hand, written lan-
guage has a structure of its own, the study of which comes
within the realm of the linguist's preoccupations, and in-
cluding the term 'vocal' in the definition of language might
be interpreted as unduly narrowing the field by excluding
areas where linguists feel perfectly competent. Besides, some
scholars always want to grab the largest possible share of the
epistemological cake. Yet, there is one important argument
in favour of including vocal nature in our definition, namely
that vocal quality is directly responsible for the linearity of
speech and the consequent linearity of script. It is clear that
if signs were visual and presented on a surface, there would
be no need for language to be manifested by a succession of
items: the painter who is presenting his message on canvas
must necessarily delineate his figures one after another, but
the recipient of the message may collect it in any order he
REALISM VERSUS FORMALISM
chooses, casting a general glance or concentrating on parti-
cular features as he sees fit. Should the message be the man is
killing a bear, onlooker might easily grasp it at first sight,
whereas the hearer will perceive it successively. The history
of writing begins with painting, i.e. a type of message which
is completely independent of language and speech, and leads
through stages to alphabetic writing which implies complete
conformity to the double articulation of language. But writ...
ing becomes distinct from painting as soon as objects and
figures are shifted ever so little from the respective positions
which they would be expected to occupy in space, in order
to suggest a succession reminiscent of the one our language
units have in speech. It is true that we could easily imagine
some other code, a gesture code or, why not?, an olfactory
code whichwould also impose successiveness. But it is hard
to imagine all the implications the particular nature of
various bearers of meaning would have had forman's com-
munication, The vocal nature of human language is cer-
tainly no peripheral aspect of it, but a basic feature, without
which linguistic organization might be fundamentally dif-
ferent from what we know it to be.
All of this points to a definition of 'language' which might
run as follows: A language is a medium of communication
according to which human experience is analysed, differently
in each community, into units (monemes) with a semantic
content and a phonic shape. This phonic shape, in its turn,
is articulated in distinctive and successive units (phonemes).
whose number in a given language is fixed and whose nature
and mutual relations also vary from language to language.
This means that we should reserve the term 'language'
for a medium of communication which is doubly articulated,
and whose outward manifestation is vocal. Apart from this
common core, nothing can be said to be linguistic which
cannot differ from one language to another. This is how we
should. understand Saussure's dictum that linguistic features
are arbitrary or conventional.
REALISM VERSUS FORMALISM
It is clear that this in spite of its unwieldy
length, does 11.0t list all the types of features that may enter
the fabric We may even be sure that it does not
include elements which probably play some role in all known
languages. There is no mention anywhere of speech melody,
which, in the minds of some educators, should embody the
most salient features and decisive aspects of the language they
teach. We might dispose of an objection coming from those
quarters by arguing that the vocal nature of speech, covered
in our definition by the word 'phonic', implies the use of
the so-called organs of speech, among which are the vocal
chords; the vocal chords necessarily vibrate at a certain
musical height, and the melodic continuum that results from
a continuous vibration of the glottis is precisely what we
referred to as speech melody, a physical reality some people
hastily identify with intonation. We shall revert to this later.
We should first answer another objection which affords
a very welcome illustration of the divergences between the
formalistic approach to linguistics and the one which t would
recommend. I must say, at once, that I do not remember
having heard it formulated, except perhaps by myself. In
any case, there is enough of a formalist left in me to anticipate
it. It is not true that any utterance can be analysed into a
neat succession ofmonemes, each with its own nicely wrapped-
up meaning and clear-cut segment. When I say he cut, where
is my segment corresponding to the meaning 'past'? In
French elle va au marche, with au a single phoneme /0/, what
slice should I ascribe to my preposition and what slice to my
definite article? As regards our second articulation, we have
seen before how difficult it is to analyse into phonematic
segments words like ice and out, and I have suggested that our
failure may reflect less the imperfection of our methods than
a factual indeterminacy.
Our answer will be that we have not said, or implied, that
the whole of language, as represented by corresponding
speech, could be exhaustively reduced to successions of
REALISM VERSUS FORMALISM
monemes and phonemes. We have said simply that what we
want to call a language makes use of monemes and pho-
nemes; whether it adds to them other tricks which may at
times blur or distort some features of double articulation is
another matter.
If we consider double articulation as the core of language,
and would see in the rest just margins, it is because this
removes language farthest from indiscriminate, unanalysed,
interjectional, prehuman, or, should we say, proto-human,
forms of communication. This alone can secure for language
the stability and rigour that result from the use of discrete
units. We all, at one time or another, may get impatient
with double articulation because some intonation or some
interjection will enable us to achieve what we want far more
quickly and at a much lower cost. But this means nothing
more than the observation that it often consumes more time
and energy to get someone to do something through asking
him than through giving him a push. Using language is a
very complex and abstract procedure which is well adapted
to a certain degree of sophistication, but which, in a number
of trivial circumstances, may economically be replaced by
some more direct means of communication such as gesturing,
either with the hands or the shoulders, or with the glottis.
In which case, if the use of language is concomitant, we
speak of intonation.
The advantage of calling intonation 'gesturing' is that it
removes it at once to the far periphery of the field oflanguage.
But of course no one will accept this, except perhaps as
a metaphor, like the Lautgebiirden of a former generation
of psycho-linguists. A gesture properly so called may be
accompanied by noise, a snapping offingers, for instance, but
it cannot be sheer noise. Only a convinced functionalist
could accept the view that a noise is not first and foremost a
noise. Since intonation is produced by the vocal chords, and
since the vocal chords are par excellence the organs of speech,
intonation cannot be anything but speech. This dictum
REALISM VERSUS FORMALISM
sounds very much like common sense, and although I, for
one, am convinced of its fallacy, I will not reject intonation
from linguistics. indeed, be made to play a role
similar to that of monemes, in such close co-operation with
the most central and abstract of linguistic tools that any
linguistic description would present wide gaps if all reference
to intonation were to be avoided.
Actually, no clarity can be achieved in our discipline
without establishing some sort of functional hierarchy. We
shall, no doubt, have to state that some feature or some aspect
of speech is not linguistic: this will be said, for instance, of the
initial rise of a speech melodic curve; this rise is due to the
fact that the glottis, starting from a state of rest, will have
to reach some degree of tension, and that the speaker is not
likely to wait until that degree is reached. Yet we shall have
to distinguish different levels of linguistic relevancy: some
French speakers pronounce their r's as a tongue-tip trill,
others as a dorsal fricative. This is found in Germany,
Holland, and in some other linguistic communities. But it is
far from universal. It is a culturally conditioned trait, not an
inescapable result ofthe working ofman's psychic and physio-
logical make-up. It has no function, if we reserve this term
to cases where some choice of the speaker is involved. It may,
however, inform the hearer about the origin, urban or rural,
of the speaker and, consequently, colour his interpretation of
the message. Some mention of all this must be included in a
description of French or German, and how could we totally
deny the epithet 'linguistic' to something which it is the duty
of the linguist to describe?
The double-articulation theory and any definition of lan-
guage based upon it leaves a wide margin, for which the
name 'prosody' is today a widespread designation. Any-
thing may be said to be prosodic that does not fit in the
monematic and phonematic segmentation; so that the Ameri-
can 'suprasegmental' is not a bad substitute. I just think
nothing is gained by speaking of 'suprasegmental phonemes'.
REALISM VERSUS FORl\1ALISM
It must, in any case, be kept in mind that some prosodic
features like intonational contours are 110t distincti;e in the
sense in which a is directly mean-
ingful: a rising interrogative contour onyou like it? has about
the same function as the do of the more traditional do you like
it? other words, it might equated with a moneme, not
with a phoneme. In Saussurian terms, it is a sign, a minimal
sign, with a signifiant, the and a signifi, 'inter-
rogation'.
The distinction should thus theoretically be made between
double articulation and prosody and not, as is traditional,
between phonematics and prosody. Yet this latter practice is
in many respects amply justified because it is very excep-
tional to find as clear-cut a signifie as the one corresponding
to the rising interrogative contour, intonation generally lack-
ing the discreteness of n1any gestures. If we consequently
agree to lump together phonematics and prosody under
phonology, we must point out that the distinction between
the two branches of phonology is based on differences in
segmentation, and that this dichotomy may at times conflict
with the classification based on function. I For instance tones , ,
or as some people call them tonemes, have exactly the same
function as phonemes: they are distinctive, which means that
the speaker, at a certain point in the message, will have to .
choose between a number of them in order to say just what he
wants to say.. It is, ofcourse, perfectly immaterial whether the
choice is conscious or not. If tones are not considered dis-
tinctive features of vocalic phonemes, it is because they are
usually found to affect, not a vowel phoneme as such, but a
syllabic nucleus, often made up of two or more phonemes
or even more than one syllable. This is shown by Swedish or
Norwegian, where the distinction between the two tones
requires at least two syllables to be preserved: kbmma 'comma'
I Regarding the methodological conflict between segmentation and func-
tion, cf. A. a r t ~ n e t 'Accent et tons', Miscellanea Phonetica, ii (London, 1954),
pp. 13-14, and Elements, pp. 54-57.
REALISM VERSUS FORMALISM 3I
and komma 'come'. In a language where syllabic nuclei
always coincide with vocalic phonemes, considering tones as
distinctive vocalic features, together degree of aperture,
lip-retraction, or nasality, might not be objectionable. But
where this is not the case, we have to separate tones and
'phonemes, although they function alike, and put the former
with accent and whatever remains ofthe melodic curves when
tones have been extracted.
We should, of course, by no means define the domain of
prosody by reference to the physical nature of the features we
want to include: should, for instance, nasality prove supraseg-
mental, we should not exclude it from prosody on the ground
that it normally appears as a phonematic characteristic; -in
many languages a glottal stop or catch functions as a tone
and is regularly regarded as such although the glottal stop
is, in many languages, a phoneme, just like [u], another
glottal product. It is, however, interesting to notice that the
features which are normally made use of in prosody are
those that are necessarily present in all utterances: stress, for
instance, conceived of as the degree of energy with which a
spoken stretch is articulated, is always there; as soon as voice
is heard, the vibrations of the glottis must have a given
frequency, which results in melodic height; duration is, of
course, unavoidable in speech, since speech exists through
time. It is therefore understandable why speakers rarely get
a chance to oppose these features to their absence at a certain
point in the utterance, but only to choose between their
modalities, which may vary from one part of the utterance
to another. Consequently, they are not so well adapted to
characterizing discrete units as others, like, for instance,
nasality or dorsal occlusion, which mayor may not appear in
a given utterance: if I say What shall we do today? I make no
distinctive use of either of these features, while I cannot help
giving my utterance a duration, using a varying amount of
energy from the first to the last phoneme, and whenever
voice is there-and voice must be there-giving a certain
I This illustration is taken from the material collected in Japan by my
colleague Robert Austerlitz.
811928 D
33
REALISM VERSUS FORMALISM
a drop is interpreted as .the end.
the lower the drop, the more final the statement will sound.
cannot be said that a rapid and deep fall 'means' con-
tempt, but it is normally indicative of some such feeling in
the speaker, or, at least, since everybody knows how to play
the game, it is what the speaker wants his audience to believe.
The less abrupt the fall, the friendlier it will sound. The
frequency of statements with level final contour in polite
British usage bears witness to an unmistakable effort not
to give offence. The least indication of a final rise will suggest
non-finality. The interrogative final rise is just one type of
those non-final contours. Within the unit of utterance itself,
the sentence, a slight rise will normally precede any pause,
since otherwise a pause might be interpreted as the end of
the sentence. In Ghiliak, there is a speech segment with the
sound [4>uru] with no specific function except the support of
a melodic rise which might be interpreted as follows: 'the
preceding pitch fall indicates the end of a sentence, but
the present rise indicates that it is not the end of the story
yet; please, don't clap your hands'. I French political oratory
makes use of a similar trick: the gasping orator, who does not
want to be interrupted by applause, lets his ,voice sink to-
wards the end of the period, but immediately tags the first
word of his next sentence with a rising pitch, which will
secure him the breathing spell he needs.
The picture we have just drawn of the main features of
speech melody is not, as it might seem, exclusively derived
from a narrow, personal experience, but is, at least, but-
tressed by recorded observation. Nevertheless, it cannot be
assumed that no language ever shows deviations from this
general pattern. In sketching it, we have not included the
possibility that some stretch of the melodic curve might be
needed for the marking of a distinction, as is frequently the
case in so-called tone languages; we have disregarded the
3
2
REALISM VERSUS FORMALISM
frequency to my glottal vibrations. The needs ofcommunica-
tion, acting on pre-existing, traditional linguistic patterns,
may yield strange results which seem to challenge articula-
tory and acoustic economy, as when, for instance, a lan-
guage with a series of glottalized consonants extends it to
the labial position, in spite of the fact that this involves using
the cheeks, which are the worst possible organs to use when
what is at stake is securing in the mouth as high a pressure
as possible. In general, however, language economy will tend
to eliminate such quirks and to restrict the play of the dif-
ferent organs to those to which they are best adapted; this
accounts for the preference evinced for stress, pitch, and
duration in the prosodic domain, and their limited employ-
ment outside of it. Within prosody, however, pitch features
seem to be well adapted to a number of different functions,
and there is perhaps no better illustration of how the same
physical fact can be used linguistically for totally different
purposes than a precise functional analysis of speech melody.
We should remember, first of all, that speech melody
cannot be left out, because voice is a normal ingredient of
speech, and voice cannot exist without pitch. Before speech
begins, organs are normally at rest, and some time will
elapse before they reach the degree of tension found to be
normal for a certain style. This accounts for the normal pitch
rise at the beginning of an utterance. Voiceless stretches,
usually very short, within an utterance do not seem to affect
the general tension of the glottis. When the end of the
utterance is getting near, the speaker will tend' to antici-
pate it and allow the organs to relax before speaking stops
altogether. This accounts for the speech drop at the end of
utterances whenever there is no reason to check it. But it must
be kept in mind that, whereas the initial rise is physiologic-
ally determined and consequently fairly stubborn, the final
drop can easily be replaced by some other contour. This
being the general conditioning, it is easy to see how variations
of that curve can be put to use in human communication:
34
REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM
35
fact that word accents may also affect the form of that curve;
we have left out of the picture the existence in nlany, and
perhaps all, communities of favoured contours which may
entail, on the part of the users, a slightly different interpreta-
tion of other contours. We mean this pattern to be nothing
but a point of reference, which should prevent describers
from forgetting how much in speech melody is conditioned
by human physiology and psychology and consequently is
not and should not be the concern of the linguist.
The first task of the linguist in prosodic matters is obviously
to spot in the melodic curve all the stretches that have
functions, in other words, the tones or tonemes. Once they
have been identified and classified, they should be abstracted
from the curve. Whether what is left, after they have been
disregarded as so many accidents, can be identified with the
above-mentioned pattern seems to depend on the importance
of tone distinction for the language, and the complexity of the
tonal pattern: in African languages with two tones, a high one
and a low one, the tonal succession does not seem to blur the
general melodic pattern, so that the latter may be relied upon
to indicate whether the utterance is a statement or a question:
'high' and 'low' mean above and below what the general
direction of the curve would lead one to expect if it were not
for distinctive tones. In Vietnamese, on the contrary, where
a six-tone system is the norm, with a high and a low register,
with one rising, one punctual, and one interrupted tone for
each, it is doubtful whether enough freedom is left to speakers
in melodic matters for them to make use of the latitude
offered by the general pattern.
If the language under description is not a tone language,
the describer will have to determine whether the language
makes use of an accent, and, if it does, how accentual pro-
minence is indicated and whether and how it affects speech
melody. Recent investigations seem to indicate that pitch
plays a much more decisive role in accentual matters than
had previously been assumed for such a language as English.
It would seem that an essential ingredient of English accent,
traditionally but perhaps mistakenly called 'stress', is a
sudden and rise. or fall of the pitch. I In any case,
accent is likely to modify the normal course ofspeech melody,
and SUCll accidents as it may cause in it will have to be
abstracted, just as tonal accidents were. A common error
consists in ascribing to a vast ill-defined domain of intona-
tion a number of features which functionally belong to that
of accent. The use of 'stress', which refers to a physical
reality, instead of 'accent' is apt to confuse even competent
scholars and make them speak of intonation as soon as they
fancy they are hearing pitch instead of stress. for many of
them, the difference between to increase and an increase would
be due to a different placing of stress, while that between
a moving van (a van used for moving furniture) and a moving
van (a van in motion) results from the use of a different into-
national contour. Once accent is defined, not in reference to
an alleged physical nature, but as prominence given to one
syllable per word, or accentual unit, with a view to marking
the respective importance of the units within the utterance,
it becomes clear that the difference between mooing van and
moving van is accentual and nothing else.
What we should call intonation is therefore what remains
of the melodic curve of speech once all tonal and accentual
features have been extracted, and, as rightly pointed out by
many scholars, most of what is of interest there centres
around final contours. It has become usual to analyse these
contours by reference to three or four levels. No one will
object to such a practice if those levels are presented and
used as a frame of reference with about the same value and
function as the vowel quadrilateral of Daniel Jones, with its
cardinal vowels in relation to which a given vocalic sound can
be characterized. It: however, those levels are meant to be
phonemes, there is bound to be some disagreement. It is,
I Cf. Dwight L. Bolinger, 'A 'Theory of Pitch Accent in English', rVord, xiv
(1958), pp. 1 9- 49.
REALISM VERSUS FORMALISM REALISM VERSUS FORMALISM
37
of course, perfectly true that if one of the levels is replaced
by some other, the meaning of the utterance may be changed,
just as the meaning of I make is changed if I replace m by t,
But, of course, m and t are phonemes because interchanging
them modifies the identity of the moneme where this takes
place, and, thus, the utterance is affected by this interchange
only indirectly through the replacement of one moneme
by another. Our problem here is whether a shift of intona-
tional level is to be equated with the replacement of one
phoneme by another or that of one moneme by another.
Let us assume, for instance, that in a contour describable in
terms of the succession of levels 2-3-2, level 3 is replaced by
level I. A peak of the curve will then be replaced by a trough.
Level 3, which, all by itself, implied a rise, will be replaced
by level I, which, all by itself too, implies a dip. This shows
that everyone of the relevant successive levels is indicative
of one of the specific directions assumed by the melodic
curve. Now, each of these successive directions .is contribut-
ing something to the significance of the total contour. This
contribution is additive, as is the one of a moneme to the
meaning of an utterance. It is not destructive, as that of
a phoneme whose presence signalizes that any inference, as
to the meaning of the utterance, which might be drawn
from the context considered without it is wrong: if, to the
statement it is good, I add very, I am just adding some addi-
tional information without deleting what was previously
there, but, if to the statement it is a roe I add a jdj phoneme,
the statement becomes it is a road; one element of information
roe is deleted and replaced by another one. Since a level is
just a way of indicating a direction, and since every succes-
sive direction would seem to add something to the whole, it
is permissible to state that the value one may wish to ascribe
to a given contour does not actually go beyond the values
that might be attributed to everyone of its successive levels.
In other words, inasmuch as the contour has a bearing upon
the sense of the message, it does so as the sum total of the
implications contributed by everyone of its components,
the levels. This amounts to saying that the relation of levels
to contour is comparable with that of words to sentence,
in which both words and sentence are signs with signifie and
signifiant, and basically different from that between phoneme
and word, or moneme. Should we insist on identifying the
items we operate with in intonational matters, with the units
that are the frame of double articulation, we would have to
say that levels are monemes, i.e, minimal units with meaning,
whereas contours are a succession of monemes. But I do not
think anything is gained through confusing the different
planes whose distinction has proved instrumental in clarify-
ing the working of language.
The central problem of intonation, a problem which,
having hardly been formulated, has, to my knowledge,
never been tackled, is that of the relation between the levels
as posited by the language describer, and their linguistic
reality, or, in other words, the behaviour of speakers, It
seems clear that they are not discrete units because it is
never quite immaterial for the message whether the 'level'
is performed a little higher or a little lower; a level implying
bitterness or contempt will be the more bitter or contemp-
tuous if it is pronounced a little lower than may be usual;
with phonemes, which are typically discrete, the precise
way a phoneme is pronounced cannot change the message:
dab remains 'dab' and nothing else, whatever the amount of
voicing of its initial jdj, as long as there is enough of it to
keep dab distinct from tab. But, if intonational levels are not
discrete, do they or do they not, in a given language, corre-
spond to some habitual comportments? If they do, it means
that a certain type of situation normally entails the choice of
a certain level, which, as we have seen, determines a certain
direction of the melodic curve. If this is the case, any per-
ceptible deviation from this direction must be considered
and classified as an individual deviation from a norm. It
is a meaningful variation, but still a variant of a type. The
REALISM VERSUS FORMALISM
I
F asked a.bout the history of structural linguistics, most
people concerned would probably say that it all began
with the phoneme. When structuralists chose first to
concentrate on what we have called the second articulation,
they certainly found the correct approach towards greater
rigour in the treatment of linguistic problems: the discrete
nature of linguistic units is ultimately based upon the dis-
creteness of the phoneme, and phonology was the foundation
vie needed for any further progress. Still, dealing so long
with phonemes before attacking the more intricate field of
the first articulation had an unfortunate consequence. having
achieved outstanding success in 'phonemics', linguists were
legitimately induced, when tackling the actually far more
complex study of the significant aspect of language, to use a
similar pattern. This accounts for frequent terminological
pairings such as phoneme-morpheme, phone-morph, allophone-
allomorph, and so forth, and, on a different level ofabstraction,
the isomorphism of the glossematicians with its strict paral-
lelism of the two planes of expression and contents. I
The most fundamental objection to this practice of identify-
ing the patterns on two different planes derives from the
obvious fact that, in language, something which is not
manifest, variously called meaning or experience, is mani-
fested by means of something else. This, no doubt, implies
a one-to-one equivalence, that of the signijiant and the
signifie, but not necessarily an identical behaviour of the
minimal, significant unit, which I call the moneme, and of
the minimal distinctive unit, the phoneme. The moneme is
s C I F s o
answer to query a mass
ofinstrumental recordings submitted to a statistical examina-
tion. This, of course, would have to be done for a large
number of the most diverse languages before a well-founded
general linguistic treatment could be presented.
That such intonational habits exist need not be further
substantiated. They are some of the features that often
enable people to state that some person has this or that
'accent'. But it remains to be determined how far such
habits actually hamper or prevent individual uses of the
natural implications of speech melody.
What, in any case, is perfectly obvious even before any
world-wide research programn1e has been carried through,
is the linguistically marginal nature of intonation, which
should be obvious as soon as a dispassionate examination has
deprived this term of its glamour and limited its application
to some residual uses of speech melody. This, to be sure, is
very much the same as to give your dog a bad name and
hang it. But need we care what terms we use, provided the
analysis is correct and fruitful?
I Cf. Jerzy Kurylowicz, 'La Notion de I'isomorphisme', Recherches structurales
(Travaux duCercle Linguistiquede Copenhague 5), 1949, pp. 48--60.
4
TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FUNCTIONAL SYNTAX
a Saussurian sign, a unit with a meaning and a phonic shape,
i.e, one which combines something that is not manifest with
its outward manifestation. It belongs to the two planes of
expression and contents, and it is the smallest segment that
does. The phoneme has a phonic shape, but no meaning.
It is pure manifestation and belongs exclusively to the plane
of expression. The moneme-and-phoneme approach to
linguistic analysis-and, for that matter, the widespread
morpheme-and-phoneme one too-does not coincide in the
least with the two-plane pattern of glossematics: considering
an utterance like give me the book, glossematicians will put,
on one side, the individual 'cenemes' like Ig/, III, [v], the
signifiants, either minimal (/gIv/, Imi/) or complex (/gIv mi
Od btrk/) , and their graphic equivalents (g, give, give me the
book) as well; on the other side the signifies, either minimal
('give') or complex ('give me the book'). The basic glosse-
matic dichotomy can be represented as follows:
Igl ) ('giVe'
IgIVI 1'.1 ,. ,
jgrv mi O;:J buk/ gIVe me the book
Linguists, who, explicitly or implicitly, operate according to
the double-articulation pattern, will keep their phonemes
alone, and put signifiants and signifies together according to
the following schema:
} {
/gIv/-'give'
jgj r-.; [gtv mi O;:J bukj-'give
me the book'
This lopsided diagram conveniently illustrates how the
moneme stands on a far higher level of complexity than the
phoneme and why any effort towards pairing them is bound
to result in distortion.
The fundamental difference between distinctive and signi-
ficant units must ultimately account for one very important
discrepancy in the syntactic comportment of phonemes and
monemes: a phoneme fulfils a function in a definite position.
If we want to identify a signifiant, e.g, that the word lake,
it is not enough to say that it is made up of three phonemes
Ill, leil, and Ikl because the same phonemes are those which
characterize the words clay and kale; one must specify: /1/,
leil, and Ikl in that order. In other words, when pronouncing
lake, speakers have to choose, in initial position, III and
oppose any inclination to say Ikl for cake, ItI for take, &c.
Postponing III till the end of the word and anticipating the
choice of Ikl would not do, because we would thus get kale
which is not what we mean. All this, which sounds trivial, is,
in fact, basic for the establishment of the phonematic pattern
of the language..
The situation is different with monemes or significant
elements generally; the relevancy oforder is far from general:
it is fairly immaterial whether I say the one I like is Paul or
Paul is the one I like; the implications are different if I say
with Paul, I went to Rome and I went to Rome with Paul,
but they do not affect the identification of the moneme
group with Paul, and the same applies to yesterday in I went . . ..
yesterday and yesterday, I went. .. . . Certainly, the respective
position of monemes is often determined by tradition or by
the need to distinguish between utterances with different
meanings: it just isn't done to say Paul with went I Rome to,
and it is far from immaterial whether I speak of root hair or
hair root, not to speak, of course, of the difference between
the mankills the bear and the bear kills the man. But it is clear that
whereas phoneme classes can be established by listing all the
phonemes that appear in a given context, this cannot be done
indiscriminately with monemes, and it might seem that the
first step with them should be to determine the situations
where their presence results from an exclusive choice, as is
normal with phonemes.
It is, however, preferable by far to forget the phonemes for
a while and try to discern what comportment we may expect
TOWARDS A FUNCTIONAL SYNTAX TOW.f-\RDS A FUNCTIONAL SYNT.A.X 43
on the part of monemes as the products of the first articula-
tion of language. Let us start again from the experience that
has to be communicated.. If the communication is to be
linguistic, that experience will have to be analysed into a
number of elements for each of which the language under
consideration has an equivalent, a moneme with its meaning
and its phonic form. For simplicity's sake we shall assume
at this point that the phonic form of every moneme is always
neatly circumscribed. The communication will thus take the
form of a succession of monemes, each corresponding to some
definite element of experience. But, of course, the choice and
nature of the elements of experience will vary from one
language to another. Let us suppose that the experience to be
conveyed could be rendered in English by means of Yesterday,
there was a riot in thevillage. One of the elements of the English
analysis is the moneme riot; riot applies to a definite happen-
ing, and it is immaterial whether it has previously been
identified or not: a riot in English is always riot, whether
it is a riot or the riot; many languages do not bother about
distinguishing between previously mentioned and not
previously mentioned; English does, and this specification
is treated as one of the elements of experience; now, we could
imagine a language which would have a different moneme
for 'a riot' and 'the riot', as if people said, for example, riot
for 'a riot' and brawl for 'the riot' .. This would, of course, be
most uneconomical, and when people really care to dis-
tinguish between definite and indefinite, they manage to
procure articles which become elements of experience in their
own right. In some languages, the simple mention of a riot
may suffice, in the absence of any restrictive moneme, to
indicate the reality of the riot; in our hypothetical language,
riot might mean not only 'a riot' but also 'there is a riot'.
English, just like a good many other languages, needs an
actualizing phrase (there was) which we might be allowed
to consider a single moneme, were it not accompanied by
a moneme indicative of time, in this case past time, This
could, of course, be easily dispensed with in an utterance
like the present one, where yesterday refers specifically to a
well-defined of past there was
a riot, with its four monemes, is the reflex of a specifically
English analysis, into four elements, of one aspect of our
experience, where some other language might get along with
only one.
Another aspect of our experience is rendered by means of
the phrase in the village. The moneme village represents, as an
element of experience, a place, but not necessarily a place
where something is happening; village, preceded, of course,
by sorne article, as here, could be made to designate the
place, or perhaps its inhabitants, as the agent of some action,
as in the village decided . . ., or as its 0 bj ect, as in they saw the
village. The function would, in both cases, be indicated by the
respective position of the elements. In our present case, the
circumstance that the village is the place where something
happened is treated as another element of experience whose
reflex is the moneme in. It would not be difficult to imagine
a language in which the moneme designating a village would,
at the same time, indicate that the village in question is the
portion of space where the experience is located. In such
a language, village would mean 'in (the) village', and this, of
course, all by itself without any case ending, since any case
ending would have to be considered the linguistic equivalent
of another element of experience.
The third aspect of our experience corresponds to yesterday.
Were it not for literary and poetic forms such asyesteryear and
the like, we might consideryesterday a single moneme since, in
that case, ... day, as the automatic accompaniment of yester-,
could not be counted as a separate choice. For our present
purposes, we shall take the liberty of disregarding yesteryear
and congeners and of treatingyesterday as a minimal unit,just
as we would do with German gestern or French hier. We have
thus, inyesterday, an exact parallel to what we have just been
imagining when we thought of a moneme village with the sense
44 TOWARDS A FUNCTIONAL SYNTAX
TOWARDS A FUNCTIONAL SYNTAX 45
of 'place where' ; yesterday, here, is not the day preceding this
day, but that day as the segment of time in which something
was happening.
We have been dividing our total experience into three
aspects, as we called them, each of which was further analysed
into monemes. From a semantic point of view, few people
would object to such an analysis. But our semantic reactions
are, to a large extent, the reflex of formal distinctions. The
formal justification of our initial break up is the fact that our
three segments there was a riot, in the village, yesterday are syn-
tactically autonomous units: everyone of them can be used
initially, medially, or finally, without any difference in their
own meaning, although, of course, the choice of this or that
order may imply som.e semantic difference for the utterance
as a whole. On the contrary, in normal contemporary Eng-
lish, word order is fixed within those segments. The reason
for the syntactic freedom enjoyed by those phrases is not
far to seek: in everyone of them we find an unambiguous
marker of its function, i.e. of its relation to the rest of the
utterance: there was marks the riot as the predicate, i.e, as the
element around which others gravitate and in relation to
which their function will be marked; in marks the village as
indicating the place where the riot occurred; yesterday, as
such, is the indication ofwhen the riot took place. The relation-
ships between the three main elements of experience are thus
precisely indicated, and there is no need here to rely on
word order to tell the hearers what these relationships are.
A language like Latin extended the practice of explicit func-
tion marking to situations where English, and western Euro-
pean languages generally, make use of word order, namely
in the indication of subject function and object function.
In Latin, a nominal subject was no part of the predicative
autonomous phrase, neither was any nominal object: pater
oidet puerum is made up of three autonomous phrases; its
English equivalent, thefather sees the child, is just one, since
the respective position of the segments is indicative of their
function, i.e, in terms of experience, their relations to one
another, which prevents their being shifted at will without
changing or impairing the message.
The criterion of syntactic autonomy points to a threefold
distinction among monemes: we have first monemes that
carry within themselves the indication of their own function
and which we shall designate as autonomous monemes:
French vite, hier, demain, dimanche in il viendra dimanche, are
autonomous monemes; in English there seem to be few
clearly autonomous monemes of that type, but autonomous
compounds such as last night, next week are at least as frequent
as in French, where we have hier soir, la semaine prochaine.
Notice, in German, autonomous compound numerals such
as neunzehnhundertneununddreissig with the meaning of 'in
1939'
Next, we have monemes that do not imply any definite
relation to the rest of the utterance and will therefore be
available for several different functions. Of course, everyone
of these functions will have to be indicated somehow, either
by position or by means of some additional element. These
monemes could be called dependants; village is a dependant.
Last, we have monemes which secure autonomy for other
monemes to which they are attached, by indicating their
function, i.e, their relation to the rest of the utterance. The
combination of such a moneme with its dependants is an
autonomous phrase. These we shall call functional monemes,
functional indicators, or just functionals. Functionals corre-
spond to prepositions and conjunctions of traditional gram-
mar, but also to case endings. The reasons for which people
are so reluctant to lump prepositions and case endings to....
gether are numerous: first, the former come before and the
latter after the form they govern; second, prepositions can be
separated from their substantives by various additions such as
an article and one or several adjectives, while case endings
are permanently glued to the word they characterize; third,
there is normally one preposition per phrase, irrespective
TOWARDS A FUNCTIONAL SYNTAX
TOWARDS A FUNCTIONAL SYNTAX 47
of how many articles or adjectives are added, whereas
case-endings are likely to be found after everyone of the addi-
tional elements; fourth, in the case of prepositions, the func-
tional usually forms a clear-cut segment of the utterance, in
contradistinction to what we find, for instance, in Latin case-
endings, where the indication of case, i.e. function, is form-
ally confused with that of a totally different type of moneme,
namely number, and where it is not always clear what
belongs to the substantive moneme and what to the ending:
is the nominative ending of puppis, 'poop', -is or just -s as in
urbs? All four reasons should not be dismissed as sheer pre-
judice. But we should never allow them to blur the functional
identity of prepositions and cases. All of them are finally
reducible to the same phenomenon: monemes, which, for
some reason or other, are frequently or constantly in contact
and will tend to merge. The merging will be the more likely
and the more intimate if the element whose function is in-
dicated comes first, and the functional indicator last. This is
due to the fact that, in any language, the number of distinct
functions is very much smaller than that of elements capable
of performing them; these elements, the so-called lexical
items, are more informative and, accordingly, generally
given a preferential treatment: they may be provided with
an accent which gives them prominence, and their initial
phonemes are, as a rule, articulated with particular care
so as to facilitate their early identification in the flow of
speech.
A formal merging of two or more monemes, I call an
amalgam: he cut, as contrasted with he admitted, can be said
to be an amalgam and to result from amalgamation. Fr. au,
in au marche, as contrasted with al' in aThopital, is also an
amalgam. We may also, if we choose, call sang, the preterite
of the verb to sing, an amalgam, although it is not likely to
have resulted from a process of amalgamation. This process
is not necessarily carried through so that it becomes im-
possible to distinguish one signifiant from another: in Slavic
languages, amalgamating processes have, at different periods,
begun to blur the boundaries between the radical moneme
ruk- 'hand' and following derivational monemes or case-
endings; this has resulted in yielding different forms for the
radical: in Czech, for instance, ruk- in the nominative singu-
lar ruka, rue- in the locative ruce, and rue- in the adjective
ruini.
Semantic amalgamation is common too, as when black-s-
mail becomes blackmail. In window, or its Danish equivalent
vindue, originally wind eye, amalgarnation, both formal and
semantic, has resulted in reducing two successive monemes
to one. When, as in a form like blackmail, the semantic amal-
gamation is complete, since, synchronically, there is no hope
of ever identifying the meaning of blackmail as the sum of the
meanings of black and mail, it is certainly advisable to con-
sider blackmail a single moneme, because, from a purely
synchronic standpoint, the homonymy of blackmail and black-s-
mail is purely accidental. But when a formal amalgam is still
identifiable as limited to a definite formal context, as is the
case with Fr. au found only before consonants and not before
vowels, or when the amalgamated forms combine monemes
with perfectly distinct functional values-as with Latin case-
endings where, for instance, nominative is a functional in-
dicator, but plural is nothing but a rnodality ofthe noun-it is
imperative to preserve the distinction. We shall speak then
of two different monemes even if an analysis of the form into
two successive segments proves arbitrary and, as in the case
of Fr. au, downright impossible. The Latin form homini con-
veys three elements of experience: 'man', 'dative', and
'singular'. Should we insist on chopping it up into successive
slices, we might rephonemicize it as /hominii/ and say that
'man' corresponds to homin-, 'dative' to -i-, and 'singular' to
-i. This analysis might be supported by a comparison with
the dative plural hominibus, where both homin- 'man' and
-i- 'dative' would reappear, -bus being then an 'allomorph
of the plural morpheme'. But, of course, hemin- means
TOWARDS A FUNCTIONAL SYNTAX
TOWARDS A FUNCTIONAL SYNTAX 49
'man' only in combination with a given type of endings,
and -i is the mark of the dative only in combination with
certain nouns in the singular. Therefore it is more accurate
and, of course, less arbitrary to say that homini means at
the same time 'man', 'dative', and 'singular' than to try
to segment it. Linguistic articulation may be blurred, and
it is our duty to describe it as we find it: we should never
try to disentangle formal units which happen to overlap or be
confused; yet we should never deny the existence of units,
as singular and plural, in Latin" whose existence is always
secured and attested by some formal difference, but in so
intricate a fashion as to defy analysis.
Our distinction of three types, autonomous monemes,
dependants, and functionals, is based upon syntactic auto-
nomy. But this leaves out one type of moneme or segment
which is independent rather than autonomous. Reverting to
our former example, we would say thatyesterday is an autono-
mous moneme, and in the village an autonomous phrase, but
there was a riot is not only syntactically autonomous, since
we can place it initially, medially, or finally at will, but
also independent, since we can use it all by itself and obtain
a complete utterance, which is not the case withyesterday, nor
with in the village. In our hypothetical language, riot, which
should mean as much as 'there is a riot', could be used all by
itself as a self-sufficient utterance. We would then say it is
used with predicative function. The situation, in our English
example, is somewhat more complex. There was a riot is defi-
nitely a predicative phrase characterized as such by what we
have called its independence. We took the liberty, in what
precedes, of considering there was an actualizing phrase and
a riot the predicate, but others might prefer another kind of
analysis and see in a riot the subject of the predicate there was.
This view is supported by the observation that an analysis
in terms of subject and predicate seems to be universally
applicable in English. Every complete utterance, in that
language, centres around a core of two monemes one of ,
which, as Sapir says, I is 'something to talk about', and
another which is what 'must be said about this subject of
discourse'. Since, in English, the subject is never an autono-
mous phrase as, for instance, in Latin when a noun, the
predicative phrase is always made up of at least a subject
and a predicative moneme, whose only possible function is
the predicative function, and which we call a verb. The sub-
ject, defined as what necessarily accom.panies the predicate,
is one function of certain classes of monemes which, in the
wake of tradition, we could call nominal and pronominal
monemes. The other functions assumed by these monemes
are those of the so-called complements. In a language where
the predicative moneme need not be actualized by means
of a nominal moneme endowed with a specific function, we
should not speak of a subject. What we would be inclined to
label 'subject' because it is rendered by a subject in a trans-
lation, is nothing but one of the complements.
We may want to define 'function' as the linguistic counter-
part of the relationship between one element of experience
and the whole of experience, so that we could speak of
function in the case of anyone of the marginal elements, but
hardly in the case of the predicative core; the predicative
function would then be no real function. This, after all, is
a matter of convention. But even if we decide to speak of
function only in reference to a definite type of relationship,
we should be ready to ascribe different functions to the pre-
dicate at least in those languages-English is one of them-
where speakers have at their disposal two or more forms of
the predicate indicating different types of relationships with
the participants of the action; if opening is the action, and the
participants a gardener and a gate, I may choose to say
either thegardener opens the gateor thegate is opened bythegardener.
We will have to distinguish between an active function, and
a passive function, these functions being, of course, nothing
but our traditional 'voices'. It is not easy to tell what the
I Language (New York, 1921), p. 126.
811928 E
5
TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FUNCTIONAL SYNTAX
functional indicators are in opens and is opened; probably ...s
for one function, is .... ed for the other. But, once again, we
should not let our analysis be too narrowly determined by
the difficulties we may experience in analysing forms.
If we leave out, as a special type, the functions of the
predicate, we should distinguish between primary functions
and non-primary functions. Primary functions are those of
elements which are directly connected with the predicate.
In a sentence like yesterday, the headofthe department dictated a
four-page letter to the secretary he had Just engaged, the four
elements yesterday, the headof the department, a four-page letter,
and to a secretary he hadJust engaged have some primary func-
tion; the function of thedepartment.four-page, hehadJust engaged
is not primary, since they are not directly connected with the
predicate dictated. Within a so-called subordinate clause such
as he hadJust engaged, the same functional hierarchy obtains
as in the so-called main clause, but we should not speak of
a predicate there, but of a 'predicatoid', and the functions of
elements directly connected with it should be considered at
best primary-like.
Among dependent monemes-those that are neither inde-
pendent, autonomous, nor function indicators-one should
distinguish the ones that assume some primary functions
from the ones whose function is not primary. The former
could be designated as primary dependents, and the latter as
marginal dependents, or determinants; in the above sentence,
headis a primary dependent, the (in the head) and department
are marginal dependents, the (in the department) is, of course,
marginal too, but marginal to the already marginal department.
Another possible distinction is that between grammatical
and lexical monemes. In order to distinguish between them,
one should set up the inventories of the monemes which are
found in specific contexts, within autonomous phrases, where
the respective position of elements is functionally relevant.
Lexical monemes are those which belong to non-limited
inventories. Grammatical monemes are those that alternate,
in given positions, with a comparatively restricted number
of other monemes. The average frequency of grammatical
monemes like of,for, with or Lat. 'genitive', 'dative', 'ablative'
is considerably higher than that of lexical monemes such
as man, rich, or eat. Functionals are grammatical monemes,
Among primary dependents some ll1ay be lexical (nouns)
and some grammatical (pronouns). Some determinants are
lexical (an adjective like great), and others grammatical (an
adjective like my, the article the, or the 'plural' moneme),
Grammatical determinants can be designated as modifiers.
Among modifiers should also be listed such grammatical
primary dependents as are part of the predicative phrase.
These include modes, tenses, aspects, and persons in so far
as their signifiant is not syntactically autonomous: in I'll do it,
I, 'll, and it are modifiers.
With the setting up of modifiers as a specific type of
moneme neatly distinguished from another type of gram...
matical element we have called functionals, we are definitely
breaking with tradition. Stressing, as we have done, the
syntactic autonomy of certain: units or phrases, is, of course,
nothing but pointing out a feature that sets apart adverbs
and adverbial phrases in so far as they are used as verbal
adjuncts. Labelling predicative monemes as 'independent'
does not go far beyond what generations of grammarians have
stated about the nature of the predicate. But dividing the
mass of grammatical elements into two basically distinct types
-the functionals and the modifiers, placing them at two far
ends of our chart, and enforcing thereby a strict segregation
-may well be felt as verging on scandaL We are so used to
listing as belonging together such categories as tense, aspect,
mood, and voice, that it may be shocking to hear someone
maintain that whereas tense, aspect, and mood are closely
akin and belong together, voice is as different from them as
different types of monemes can be. We are used to consider....
ing gender, number, and case as the three pillars on which
any decent nominal system is built, and we may be tempted
52 TOWARDS A FUNCTIONAL SYNTAX
to any that has no
and that number is a local accident of limited importance,
whereas case belongs to the constitutive frame ofany utterance.
Yet the distinction between functionals and modifiers is
fundamental, and it has lately been noticed and pointed out
in various quarters. I If in a phrase such as with a smile, the
primary dependent smileis considered the centre of the phrase,
the grammatical determinant a is centripetal, the functional
with centrifugal: a is connected with the rest of the sentence
only through smile, which it helps to specify; with connects
smile with the rest of the sentence, and since the connexion
is thereby established, the speaker is free to place the phrase
with a smile anywhere he pleases. Syntactic autonomy is thus
the criterion which, in all cases, and particularly in formally
complex ones, will prove the presence or the absence of a
functional. In a context like the hunter was killing a bear with
his spear, neither the nor a are function.als, since they do not
grant hunter or bear any syntactic autonomy: exchanging the
place of thehunter and a bear will result in conveying a totally
different experience; with, on the contrary, makes it possible
to place the phrase with his spear practically anywhere with-
out changing its relation to the rest of the sentence. Whether
this or that functional mayor may not be used is partially
determined by the communicative needs at that point of the
utterance: after he is distributing tickets, I mayor may not
specify to whom. But it also depends on which predicative
moneme I choose, whether I mayor may not use a to-com-
plement. It is what is referred to when we say that a given
verb governs this or that case. This amounts to saying that,
to a large extent, the choice of a functional is predetermined
by that of the verb of the clauseo On the contrary, the choice
of the modifier is free, i.e, the speaker is determined to use a or
the, the plural or the singular, at a certain point by direct
I e.g. by Richard S. Pittman, starting with A Grammar of Tetelcingo (lvlorelos)
Nahuatl (Language Diss., No. 50) (Baltimore, 1954), pp. 6-8, and Georges
Gougenheim, 'Morphologie et fonctions grammaticales', Journal de psychologic
(1959), pp. 417-26.
TOWARDS A FlTNCTIONAL SYNT.A.X 53
reference to what the experience really is that he wants to
communicate. If: in some cases, the range of the choice hap-
pens to be limited, it will be so on account of the nature of the
unit that is modified, as when a certain noun is never used
in the plural, but not because of some limitation imposed
by the structure of the sentence as a whole.
In spite of such fundamental differences, functionals and
modifiers have so far generally been confused, and it is easy
to why linguists who, in their overwhelming
majority, had been taught Latin grammar first, were
tempted, at best, to consider them two aspects of the same
linguistic reality. In Latin, as in all Indo-European languages
that have preserved the old declensions, the expression of
case and that of number are hopelessly mixed under the form
of case-endings that resist any formal analysis into successive
segments. In other words, amalgams of a functional and one
modifier, or more than one, are practically the rule there,
and must have been felt by generations of linguists to be a
normal feature of any self-respecting language structure..
As a matter of fact, it cannot be considered a strange quirk
of Indo-European, because amalgams of functionals and
modifiers are by no means absent elsewhere. It is quite clear
that the signifiants of both are likely to appear in close
succession in the utterance and are therefore constantly
exposed to amalgamation.
But the existence of such amalgams would not suffice to
explain the traditional disregard of the difference between
centrifugal and centripetal grammatical elements: after all,
the existence of complex case-endings never prevented gram-
marians from distinguishing between case and number.. TIle
main reason why the distinction was not made lay in the
existence of concord. It was generally believed that the justi-
fication for the somewhat clumsy repetitions which charac-
terize this phenomenon was to be found in the way it
allowed the speaker to connect the successive elements of the
utterance. This was professed even by a scholar like Otto
54
TOWARDS A FUNCTIONAL SYNTAX
TOWARDS A FUNCTIONAL SYNTAX
55
had patience with roundabout ways of
expressing oneself and denounced concord as a most awkward
approach to sentence building. Now, concord which amounts
to using several discontinuous segments for one and the
same signifie is attested for functionals and modifiers alike.
This circumstance was, of course, one more reason for identi-
fying them. But what is really basic here is the fact that,
thereby, modifiers would seem to be made to function as
connective elements, a role which we have been denying
them and entrusting exclusively to functionals, The answer
to this is that, ifwhat is normallyjust a modifier, redundantly
and discontinuously expressed in different positions through
the utterance, happens to act as the sole indicator of a
function, it has to be interpreted, when this is the case, as
an amalgam combining modification of the accompanying
monemes and indication of their relationship to the rest of the
sentence: in a Latin context like viri vident, viri contains
a modifier, the 'plural' moneme, and a case, the nominative
which marks its subject function; in vident the final -nt is
exclusively a part of the signifiant of the 'plural' moneme we
have just identified in viri. It is worth remarking that the
amalgam of the two monemes 'plural' and 'nominative' is
only partial, since it could be said that the former's signifiant
is -i ... -nt, while the latter's is just -i. But if our subject
belongs to a type where the nominative ending is identical
with the accusative ending, if, for example, our context is
homines animal vident, the -nt of vident becomes instrumental in
identifying homines as a nominative, since it indicates that the
subject is in the plural and, in this context, points to homines
as the only form which can be interpreted as a plural subject.
Thus, owing to the ambiguity of the nominal ending, the -nt
ofthe accompanying verb may be made to play the role ofthe
elements of a discontinuous moneme indicative of the sub-
ject function of the neighbouring plural noun. All this may
sound rather tricky, but the facts themselves are so, not our
interpretation, and this is something to which any schoolboy
would readily bear witness .. Everything would be simpler
if the nominative case were always unambiguously distin-
guished from the other cases. There then never be
any need to resort to the mark of the plural agreement to
indicate which noun is the subject. The intricacies of Latin
are no proof that the distinction between functionals and
modifiers is not fundamental,
Concord is redundancy, and contrary to what could be
expected, redundancy results, as a rule, from least effort:
people do not mind repeating if mental effort is thereby
reduced; if adjectives are quite freely and frequently used as
nouns, as was the case in older Indo-European languages,
it will be indispensable for them to carry the mark of their
function if nouns do; therefore a word likefortis, 'courageous'
or 'courageous fellow' is inflected just like civis, 'citizen'. If I
mean something like 'the citizen, the courageous one', there
is some justification in presenting the mark of their function
twice, since the two words are more or less equals. But when
fortis is really nothing but an attribute of civis, it would be
more logical either to mark it as such, or to use the bare
stem, letting its vicinity to civis indicate its relation to it. But
it is certainly much easier to let the adjectives keep in all
cases inflected forms that have had to be memorized any...
way. In popular French the equivalent of lnyfather saysis mon
pere it dit (pronounced [idi]) instead of the traditional mon
pere dit; since I have to use [idi] in reference to my father's
saying something when I need not specify that the speaker
is my father, why shouldn't I save myself the trouble of
choosing between [di] and [idi] depending on whether I
expressly mention my father or not? Since I can't help using
[idi] at times, it is handier to stick to it, whatever the context,
than to reckon every time with the context; pronouncing an
extra phoneme is nothing in comparison with the output of
mental energy required by the choice between [di] and [idi].
There are situations where it is not usual to speak of con-
cord or agreement, but in which speakers also have recourse
TOWARDS A FUNCTIONAL SYNTAX
TOWARDS A FUNCTIONAL SYNTAX
57
to redundancy for obvious reasons of economy: in our
former Yesterday, there was a riot, the notion of 'past' is
expressed once precisely, inyesterday, and then rather
vaguely, in the preterite was. Now, it may be useful to have
at one's disposal, together with a set of references to some
definite periods of the past, such as yesterday or last week,
some means of labelling the experience as a past one without
any further specification. Of course, if we start with a precise
reference such as yesterday, there is no need to add a vague
reference to the past as the one included in was. The lazy
solution, however, consists in letting the context, yesterday,
determine the choice of the tense. In similar cases, one might
easily be tempted to characterize the tense moneme of was
as relational, since it sluggishly- establishes some connexion
between different parts of the utterance. In a similar way, it
could be argued that such a modifier as the definite article
implies some relation with what precedes, since it normally
labels its noun as something previously mentioned. Such
facts are what Sapir probably had ill mind when he labelled
all grammatical items as relational. I
It has now become clear that redundancy, grammatical-
ized or just lexical, is a basic feature of human communica-
tion, and it may in particular instances be resorted to in
order to indicate the function of some segment. But, once the
uses of the respective positions of monemes have been dis-
counted, only such elements as secure syntactic autonomy
are to be considered functional indicators. Whether this
syntactic autonomy is frequently or exceptionally made use
of is immaterial. Even if it could only be shown by means of
some unidiomatic syntactic shift, it still would have to be
deemed valid, if it were proved thereby that the identifica-
tion of functions is not affected.
Most of the preceding examples were borrowed from
English, with a sprinkling from other genetically connected
lOp. cit., pp. 86-92.
and structurally similar languages. This might lead readers
to believe that our analysis is based upon the observation of
a certain type of language and, that it cannot
make any claim to universal validity. But this assumption
would not be correct: starting from our definition of language
as doubly articulated, we have actually been proceeding in
a purely deductive way, and the existing forms we have been
quoting and even operating with were never meant as sup-
port for the theory, but simply as illustrations ofthe categories
we posited as necessary for the functioning of linguistic com-
munication. What has led us to stress the importance of
syntactic autonomy is the realization that this autonomy is
the test that a segment of an utterance corresponding to a
given element of the experience contains all that is needed for
marking its relation to the rest. It is a guarantee that the
hearers will be in a position to reverse the process through
which the speaker breaks down his previously unanalysed
experience into a number of elements for which the language
he uses offers equivalents. It would seem that there exist only
three ways for the speaker to indicate the relation of a seg-
ment to the whole:
First, the linguistic equivalents of the elements of experi..
ence may be connected by means of units, in all respects
similar to them, which we might consider new elements of
experience in their own right: in John's hat, 's is but a handy
way of expressing 'belonging to' which is part of the experi-
ence just like 'John' or 'hat' ; in the same way as some elements
of experience may be represented by prosodic features, the
marking of relations may be secured by means of an intona-
tional or accentual feature"
Second, the relation between two elements may be ex-
pressed by means of the respective positions of their linguistic
equivalents in the speech continuum: in many languages,
a subject is marked as such because of its position before the
predicate; in John's hat the respective position of John and
's marks John as the owner"
TOWARDS A FUNCTIONAL SYNTAX TOWARDS A FUNCTIONAL SYNTAX 59
Third, the relation of an element to the experience as a
whole may be included in the 'meaning' of its linguistic
equivalent: yesterday corresponds to an element of experience
whose relationship to the whole is never in doubt. But, beside
such perfectly autonomous segments as yesterday, we find
cases where the 'meaning' of the moneme may, in certain
contexts, imply a given function, as when Russian stol 'table',
as an inanimate masculine, will tend to be considered a
grammatical object, in the absence of any specific indication
of that function.
Proceeding in a strictly deductive way, it is our duty to
determine as we have just tried to do, what possibilities are
afforded by the linear form of speech for the linguistic ex-
pression ofthe various functions corresponding to the relations
assumed to exist among the elements of experience. But we
have no right to posit the existence of relational universals:
offhand, we may be inclined to believe that the with-type of
relation, or with-function, exists in all languages ; but observa-
tion reveals that, in many languages, the equivalents of do it
with a hammer and he came with afriend make use of different
functionals; even in a language like French, wherefais-le avec
un marteau and il est venu avec un ami show the same equivalent
of with, there are so many specific contexts where avec does
not correspond to with that it would not be commendable to
equate their respective functions. Even on a more funda-
mental level, we should be wary of following Sapir when he
presents the subject-predicate pair as the necessary basis of
linguistic communication. I Here again, we have to envisage
various possibilities and try to imagine what could constitute
a minimal utterance, and how such an utterance could be
further expanded.
The first problem that arises in connexion with the minimal
utterance is whether we should distinguish between normal
utterances, the ones which, in English, involve a subject and
a predicate, and curtailed ones such as over there! the scoundrel!
lOp. cit., p. 126.
fire!, not to speak of monomonematic or polymonematic in-
junctions such as go!do! getaway! give himashilling! Traditional
grammarians do not any difficulty in such matters
because it is obvious for them that, once injunctions are set
aside, only subject-predicate cores can be dubbed 'normal'.
But this is precisely what we do not accept, because we want
to consider the possibility of languages using utterances
made of one (predicative) moneme in exactly the same situa-
tions where other languages, such as English, cannot dispense
with the complex subject-predicate construction.
It is by .rcference to our distinction between what is
properly linguistic and what is marginally SOl that we may
hope to give a sound foundation to the concept of 'normal
utterance'. What we consider properly linguistic is what
is achieved, in matters of communication, by means of the
double articulation pattern: double articulation is what pro-
tects the linguistic frame against interference from outside,
what makes it really independent and self-contained. But
. communication by means of double articulation is an expen-
sive procedure which man will tend to avoid when his needs
can be satisfied through the use of simpler, more direct
means, such as gestures, by themselves or supplemented by
speech. Another, very effective, way of reducing the output
of energy involved in communication is to rely on the situa-
tion in which the interlocutors are placed: very nice! pooh-pooh!
no! make excellent sense all by themselves among people
who look at the same thing or witness the same event. This
reliance on situation is so general that all languages have
developed several classes of monemes whose interpretation
is always dependent on situation. Such are demonstratives
like this, that, except when used in reference to context, time
references like now,yesterday, today, last night or the 'preterite'
moneme, and personal pronouns like I andyou. The situation
generally makes it so obvious who the second person subject
of imperatives is, that its expression is the exception rather
I See above, pp. 28-29.
60 TOvVARDS A FUNCTIONAL SYN1'AX TOWARDS A FUNCTIONAL SYNTAX 61
than the rule. All these economical tricks are very welcome
in linguistic practice, but they undoubtedly detract from
the ideal of human communication, is self-sufficiency.
This ideal finds its expression in out-of-situation uses of
language as realized, e.g. in gossip, where the reference is to
absent people, in narratives generally, and in literature,
where use is seldom made of the actual situation, namely
the author at his typewriter and the reader with his printed
pages and, in between, the long-drawn processes of editing
and composition. It is true that authors do create situations
where their characters are found to use, with perfect rele-
vancy, such terms as I, you, today, or this week. But these
situations are actually contexts, and therefore a purely
linguistic accompaniment.
It .. is in reference to out-of-situation uses of language that
normal syntax can be defined : normal syntax is that which
is attested in such uses, i.e. when communication is achieved
by purely linguistic means. This should, of course, not be
construed as excluding from normal syntax such segments as
I tell you: it is clear that I tell you, as a syntactic pattern, is
identical with they tell them, which Inay contain no reference
to situation; in other words, I tell you is normal syntax, in
spite of the fact that this segment contains references to
situation, since it conforms to patterns attested for segments
that do not contain SUCll references.
This situation criterion is certainly valid for determining
what could be called syntactic normalcy in a language, like
English, where it agrees with the consensus of generations of
grammarians that syntactically normal utterances contain
a subject-predicate phrase. application to languages that
we suspect do not demand a bimonematic core may not be
quite as decisive: there are many languages where a moneme
meaning 'rain' (with no possibility of deciding whether verb
or noun) is used by itself when the English say it is raining.
In such an utterance, there is no linguistic reference to any
situation; it usually indicates that it is raining here and now,
but the 'here and now', corresponding to no linguistic units,
are no parts of the communication. Still it could be argued
that and now' is in the of the
moneme for 'rain' unless specifically excluded by the addi-
tion of some complement like 'yesterday' or 'beyond the
hill' . Yet, there is no doubt that languages in which the
normal way of saying it is raining or there is afox is by means
of a single moneme 'fox', deserve not to
be classed indiscriminately .with languages in which this is
not possible. The more so if it can be shown that this lTIOnO-
monematic utterance may be normally expanded by means
of various complements, whereby it is revealed as the
potential core of an unlimited syntactic complex '[here
comes a] fox' > '[there came a] fox, last year, that was killed'.
Even if the minimal utterance in a language cannot be
unambiguously shown to coincide with a single monerne, one
should not jump to the conclusion that it must necessarily
be one of the subject-predicate type. It is easy to understand
why so many languages have made it a rule never to use one
moneme by itself: even when centring his attention on the
existence of a single being, thing, or process, a speaker will
normally not be satisfied with the mere mention of that item,
but will be inclined to locate it in time or space, or to connect
it with himself or his interlocutors. These additions are
often conceived as actualizers, which seems to imply that the
moneme by itselfis an abstraction whose anchoring in reality
can only be achieved by means of some element endowed
with just that function. This view is supported by the nature
of the subject, which is a moneme, likely to be found else-
where in a variety of complemental functions, used here as
the necessary accompaniment of a predicate, with a function
perfectly characterized either by some privileged position
as in English, or some functional mark as in Latin. But the
fact that some languages make the actualization of the predi-
cate a definite function, does not imply that actualization is,
in principle, more than one aspect of the basic linguistic
62 TOWARDS A FUNCTIONAL SYNTAX
process according to which communication can be made
more specific by means of additional elements. A distinction
may be made between optional specification and compul-
sory addition, as of a subject, for which the term 'actualiza-
tion' could by convention be reserved. But it should be kept
in mind thata compulsory subject does not really actualize
a predicate more than an optional complement would.
There may be languages in which one-moneme utterances
cannot be considered normal out-of-situation syntax, but in
which the predicate can be actualized by means of any com-
plement, the actualizing function being added to any other
function the complement or complements may assume.
Traditionally, the term 'predicate' is defined in reference
to the subject-predicate complex and would seem to designate
everything in the clause that is not the subject, or some
dependent of the subject. Besides, 'predicate' implies some
assertion, so that a question or an order would not contain
any predicate. If the term is to be retained by contemporary
linguistics, we shall want to use it in reference to mono-
monematic segments which, by themselves, may constitute
a complete, out-of-situation utterance, and also to the same
segments when accompanied by various expansions (com-
plements), but independently of them. Within the subject-
predicate complex, 'predicate' should, in a similar fashion,
apply to any segment that is, jointly with the subject, con-
stitutive of the minimal utterance, thus excluding from it its
various complements.
A subject is different from a complement only because it is
constitutive of the minimal utterance. Therefore, whenever
we speak of a subject we are referring to a linguistic situation
in which subject and predicate are both equally indispens-
able, since our criterion is indispensability, and the problem
arises of how we can tell what is the one and what is the
other. Ifour terminology makes any sense, the subject should
be the one which should somehow stand closer to the
marginal and optional elements of the utterance: in many
TOWARDS A FUNCTIONAL SYNTAX 63
languages, English among them, subject function can be
ascribed to such monemes (nominals) as are found elsewhere
to assume the functions of complements ; in some, such as
Malagasy, monemes with subject function must be marked as
previously known or mentioned, which shows them to be
informationally marginaL The possibility that some languages
do 110t clearly distinguish between two successive statements
and the succession subject +predicate cannot be ruled out.
In languages where actualization is needed but can be
achieved by means of any complement, the predicate is
obviously the one moneme which is not marked as perform-
ing any complement function.
All this, and what could be added about the different
levels of complementation, amounts to establishing a hier-
archy of syntactic functions. This hierarchy is, no doubt, set
up with a view to differences between languages, but it
should be clear that, even ifit were carried through, it would
never account for all the varieties of linguistic structure.
It is quite essential to know all the different functions that
characterize a language, but it is equally important to deter-
mine for each language, what monemes are qualified to per-
form this or that function. No language is known to allow
everyone ofits significant units to perform all of the functions
it provides. It is even difficult to imagine how such a language
would work. II: in such a language, functions should be
marked by means of functional indicators, these, being signi-
ficant units in their own right, should be able to assume the
role of functioning elements, and, conversely, all functioning
elements would also have the role of functional indicators.
But how could speakers make clear that a given moneme
is used as a functioning element here, and a functional
indicator there, ifonly the respective position of the monemes
in the speech continuum could be relied upon? Many
languages are recorded in which the same moneme is used
either as a functioning element with the meaning of 'to give'
or as a functional indicator with a dative value. But in a
64 TOWA R D S A FUNCT IONA L SY N T A X
language like Vietnamese, where this is the case, not all
'verbs' function as 'prepositions', nor vice versa, and this
determines significant contexts that enable the hearer to
identify 'give' as a predicate or as a dative function marker.
There are languages, and Vietnamese is again a case in point,
in which position plays a great role in function marking; the
main function here can be labelled 'determination', and this
is shown by postposition of the determinant; what looks like
a subject-predicate relationship might possibly be inter-
preted as a case of determination. But if the core A receives
two determinants, Band C, how can the hearer know that,
in the succession C is not a determinant of the im-
mediately preceding B, but one of concomitantly with B?
Some marker, segmental or prosodic, will be needed, and
there goes our monematic omnivalence!
A language in which all monemes would be of theyesterday
type, i.e, syntactically autonomous, because the indication of
the function is part of the meaning of the term, would be so
uneconomical that we may as well rule it out as a practical
impossibility. Function being, in such a hypothetical case,
diluted in lexical meaning, would, of course, cease to exist as
such.
We are thus induced to accept, at least as a pragmatic
assumption, the view that there exists in all languages some
distinction between monemes as regards the extent to which
they may assume the various existing functions. In no
language are all monemes used indiscriminately as function-
endowed and function-marking. In other words, there is no
language without grammar. But once unambiguous function-
marking is secured, there is no universally valid reason why
any moneme, except one that is specifically a function-
marker, should be excluded from any function, whether
predicative or non-predicative, Still, specialization is very
widespread. Many languages have, for example, a class of
adjectival monemes more or less restricted to certain specific
functions and which, accordingly, tend to identify meaning
TOWARDS A FUNCTIONAL SYNTAX 65
and function: small, for instance, implies not only 'smallness',
but also the function of determination, as in a small car, and
also in he is small if we agree to locate the predication is.
It is even more frequent to find that a class of monemes
(verbs) is restricted to predicative function, although this
does not necessarily imply that only verbs can be predicates.
Contrary to what was generally assumed, this is a domain
where languages are found to vary and which, in consequence,
should playa great role in typological matters.
81192B
F
III
LINGUIS r o TYP L
W
H AT contemporary linguists somewhat pompously
call 'typology' is not basically different from what
a long line of thinkers have attempted to do when
they classified languages, not according to their antecedents
and genealogy, but with respect to their directly
characteristics. It is not difficult to understand why so Iittle
has been achieved along these lines to date: for more than
a century, practically all responsible and reputable linguists
were engaged in looking for traces of linguistic kinship, i.e.
for evidence pointing to the fact that different languages were
divergent forms of one and the same older form of speech.
We need not go too far back in time to find and, as far as
I am concerned, to remember a period when attempts to
classify languages on any other basis than genetic relation-
ship were frowned upon as a sheer waste of time and energy.
Let me only mention the fact that the second edition of Les
Langues du monde.' published as recently as 1952, has retained
genealogy as the well-nigh exclusive principle of classifica-
tion. This, no doubt an extreme case, obviously results from
lack of information, on the part of the surviving editor,
concerning the mid-century status of descriptive linguistics.
But the well-known reluctance of Meillet for any non-genetic
comparison was perfectly justified at a time when there was
no proof yet that linguistic description could be performed
with as rigorous or more rigorous methods than the ones
available for genetic comparison. Although Sapir's typology,"
I The first edition, published in 1924 in Paris, was entitled Les Langues du
monde, par un groupe de linguistes, sousla de A. et Cohen.
2 As Chapter VI ('Types of Linguistic Structures) of hIS ,?ook Langua?e
(New York, 1921), pp. but it.is based upon the analy'sls presented
Chapter V of the same book ( Form In Language; Grammatical Concepts),
LINGUISTIC TYPOLOGY
.... 't-,,,,..... and
still must be considered the necessary basis of reference for
all typological attempts, we may well understand why so
many linguists remained sceptical and why it has never been
used.
Our fundamental objection to previous attempts at classi-
fication is, of course, that they were bound to be random
and intuitive, since so little was really known about the
languages which we so glibly labelled isolating, agglutinative,
or inflexional. If many linguists are convinced that it is high
time to tackle typological problems, it is because they think
that we are in a position to give a generally valid description
of any language, one which is not biased by the background
and previous experience of the describer. I Furthermore,
they deem that we have at our disposal reliable descriptions
of a good many languages, and, last but not least, that we
know exactly what we can expect from a language, because
we ourselves have decided what we want to call a language.
This does not entail, by any means, that reliable descrip-
tions will give us the clue to the problem of establishing a
typology which will be found acceptable by anybody but the
one who sets it up. If so little has been done about typology
during the last decades, it is because no one knows how to
establish a hierarchy among the various items isolated by
linguistic analysis. Sapir, it is true, taught us that we should
not try to reduce the complexities of a language to a single
label.> But if we try to do justice to all the features that seem
to deserve mention, we may end up, not with what we ex-
pected to obtain, namely a number of classes among which
pp. 86-126. J. Greenberg in 'A Quantitative Approach to the Morphological
Typology of Language', I]AL, xxvi (1960), pp. 178-94, translates Sapir's
scheme into currently fashionable jargon.
I In this respect, Charles Bazell, Linguistic Typology, An Inaugural Lecture
Delivered on 26 February I958 (London, 1958), seems unduly pessimistic, parti-
cularly in connexion with the phonological level (p. 5); actual descriptions,
as distinct from programmatic illustrations of a descriptive method at the
expense of a given language, manage to present traits such as everyone con-
cerned would agree to consider relevant. 2 Op. cit., p. 128.
68 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 6
9
1..,.",nr"1"1..,.nrt:l.('l could be distributed, but with a long
list of criteria which, when applied to a language, would
probably set it apart from all others. In other words, what we
could achieve would be a method for trimming our descrip-
tions to their essentials rather than a handy tool for rapid
orientation.
Linguistic analysis yields, for each language, a most com-
plex picture with a list of phonemes, a prosodical pattern,
a set of morphological alternations, a grammar which some
would still divide into a morphology and a syntax, and a
lexicon, a kind of rambling and unstable structure with which
linguists do not know too well how to operate. The ideal
solution would be to reduce all these aspects to one spectrum.
In order to achieve this, we might be willing to sacrifice
certain features felt to be of secondary importance. But,
even so, it is doubtful whether we could ever succeed:
linguists have lately been repeating that a language is a
structure, or, maybe, a structure of structures, and if this is
true, we should expect to find inner connexions extending
from one end of the complex to the other end. This would be
true if a language were one of those tools or machines that
work with perfect accuracy and without any appreciable:
delay in transmissions. But this is not the case. As we shall
see later, I every language retains features which result from its
functioning several millennia ago: French il est, ils sont, German
er ist, siesind, preserve to this day an alternation determined,
in Proto-Indo-European, by the accentual pattern that must
have characterized the verbal inflexions at a remote stage of
that language. Redundancy, a normal and necessary feature
of all language, constantly acts as a buffer, retarding reac-
tions to such an extent that it may take thousands ofyears for
all the implications of one push to spread to all parts of the
structure.
The question has often been raised, whether there was any
solidarity between the phonological pattern of a language
I See below, p. 138.
and its grammatical structure. what is meant thereby is
that a change on one of these two planes is likely to have
repercussions on the there is ample proof that such
a solidarity exists: the connexion between umlaut and Ger-
manic morphology is obvious, and umlaut as a phonological
shift has resulted in umlaut as a morphological device as
used, for instance, for the forming of new plurals. Yet, if it
is true that any phonological change may involve some
morphological change, it is also true that the nature of the
phonological change will not determine the nature of the
corresponding morphological change: the confusion of e and
a may, in a given language, entail the confusion of the indica-
tive and the subjunctive; but this is no indication that there
should be a natural and permanent connexion between the
preservation of e and a as distinct phonemes and the reten-
tion of a distinction between the indicative and the sub-
junctive. On a strictly synchronic level, it is likely that, while
the morphology and the lexicon of a language require the
existence of a number of phonological distinctions, what
those distinctions actually are is of little or no importance.
This amounts to saying that it is hard to imagine how a
linguist who knows the phon.ological pattern of a language
could guess what the main features of its grammar are or,
vice versa, how, having been told what the grammatical
categories of a language are, one could tell what its phonemes
must look like.. It is clear that if the latter were the case we
need not worry about the role phonology should play in our
typology, since the essential of it should be deducible from
a consideration of the morpho-syntactic categories. In fact,
however, phonology will have to be represented in a linguistic
typology, unless we decide that what it has to offer has no
bearing upon what our typology is meant to elucidate.
It is a fact that those who have tried to classify languages
so far have not attempted to make use of phonological
criteria, unless they were professional phonologists who, in
their turn, were not trying to go beyond phonematic or
LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY
7
1
prosodic patterns.. Traditionally, classifiers concentrate on
morphological traits: they probably believe a linguistic
classification should first and foremost at grouping the
languages of people who share the same outlook.. Now, they
deem the phonology of a language to a matter of chance,
and the lexicon to reflect the world as it But its morpho-
logy, which ultimately governs its syntax, is supposed to
mirror the psychic activity of its users, When scholars
distinguished among isolating, agglutinative, and inflective
languages, they certainly believed they were transcending
mere linguistic reality and reaching deeper into the psyches
of speakers.. Some of them-e-but should we call them 'SCllO-
lars' ?-even went so far as to connect isolation, agglutina-
tion, and inflexion with some fundamental, permanent, and
inherited features of different strains of men, as when one of
them declared that it was a crime for an inflecting woman to
marry an agglutinating man.'
All this trend of reasoning is, of course, hopelessly out-
dated, less because of a weakening of racial prejudice
among scientists, than on account of the widespread con-
viction that language is more than the reflection of the world
in the mind of man.. There is something we have a right to
call linguistic reality which combines signifie and signifiant
and which is distinct from both the phonic and the mental
realities. Language is an institution. A language is a set of
habits that the child-or the adult in the case of a second
language-acquires by imitation of those who surround
him. If we want a linguistic typology today, it is certainly
not with a view to establishing a hierarchy of human races
or discovering distinct varieties of human minds.
This leads to the question: for what purpose do we want a
linguistic typology? It would probably be more truly scienti-
fic not to consider ulterior motives. At the present time,
as applied linguistics gains more and more importance, it
is imperative to insist on the necessity of promoting free
I Quoted from Sapir, op. cit., p. 131, n. 2.
research with no other aim than the expansion of knowledge..
Still, we cannot work here without criteria for determining
is what is essential. These criteria
cannot be chosen arbitrarily if we want to reach agreement
among linguists; they are likely to be generally accepted only
if the final result opens new vistas and reveals new fields for
research. This means that we should try to determine the
possible reasons why two languages which are not genetically
related, i.e, derived from one and the same language and
therefore likely to preserve substantial likenesses, should
belong to the same type.
When we come across two languages which present striking
resemblances and which, for some good reason, we do not
believe to be genetically related, we are inclined to assume
a process of convergence determined either by protracted
contacts between two communities or by some common
substratum. When we have to deal with a number of lan-
guages which we suspect to have converged, it may prove
useful to operate with the sort of concentrated characteriza-
tion which we call a typology. It: for instance, we consider
the languages spoken in the northern half of Asia, which
certainly belong to different stocks, we may imagine how
handy a reliable way of characterizing languages would
prove for those who try to unravel the linguistic history of
that part of the world through an evaluation of the duration
of convergence. or divergence. If this is the kind of use we
have in mind when considering a linguistic typology, it is
clear that phonological characteristics art likely to be just
as informative and essential as any of those selected from
grammar or lexicon: in a non-genetically orientated com-
parison of the languages of South Africa, the clicks of Zulu
should playa major role.
It is, however, frequently assumed that convergence is
likely to breed purely outward resemblances, the ones that
result from direct imitation of some sounds, or from borrow-
ing of lexioal elements or loose grammatical items, which,
LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY
73
since they are not welded to radicals, are easiest to isolate
and to transfer from one language to another. There is some
truth this, although I am of the opinion that there is no
limitation to the extent to which two languages can con-
verge. Convergence will show in trimmings before it mani-
fests itself in fundamentals. Therefore, if similarity is found
exclusively in the structural cores of the two languages, we
may be induced, when genetic relationship is ruled out, to
reject as an explanation the sort of convergence that results
from contacts. We may prefer to work with a theory which
has division of labour and increasing social complexity as the
.JI1Clin factors of linguistic evolution! so that, if by any chance
their rhythm happens to be the same in two distinct com-
munities, linguistic evolution may follow very much the
same tracks in both. We might, for instance, assume that
on a certain cultural and economic level we stand a good
chance of coming across what has been called the ergative
construction." On another level, which we might consider
more advanced, the ergative construction will probably have
given way to an active-passive verbal opposition.
Now, ifsuch a theory should be found to tally with at least
part of the observed facts, it would entail a typology based
upon such linguistic features as are likely to be most directly
affected by social and economic changes: these obviously
include lexical items, but also the elements that I would be
tempted to call the joints, and which are the essential parts
of grammar.. This would certainly strengthen the traditional
view according to which a non-genetic classification of
languages should be based upon morpho-syntactic traits.
Even if we grant that the incentives to linguistic changes are
to be found among human communicative needs, still changes
will not take place irrespective of the existing vocal forms,
and the distinctive units of the language will prove an effec-
l[ See below, pp. 136- 8.
2 Cf. A. Martinet, 'La Construction ergative et les structures elementaires
de I' enonce', Journal depsychologic normale et pathologique, 1958, pp. 377-92
tive bulwark against random distortion.. t But communica-
tion begins with a first analysis of experience in the frame of
the morpho-syntactic of language, and the way
this first analysis is performed seems far more fundamental
than the materials which are made use of for its transmission
to other members of the community. Ifcarried through to its
ultimate consequences, this statement implies that not only
phonemes, but also the vocal form of meaningful units
should, in a linguistic typology, be subordinated to the
articulations of experience considered apart from the signi-
fiants which are needed for their transmission: what is essen-
tial about the plural of English is not the choice of the
[z] phoneme for its most usual rendering; it is not the fact
that [z] alternates with lsI and /IZ/ in certain phonic situa-
tions; it is not either its being eked out by /dn/ in oxen, a
vocalic alternation in men, zero in sheep or deer, all this being
buttressed by verbal agreement with the subject of the clause.
These circumstances can all be lumped together as formal
accidents. No doubt, they may ultimately be instrumental
in shaping the fate of the plural category. But, from a
synchronic standpoint, they are just side issues in comparison
with the basic fact that English, unlike some other languages,
distinguishes, for certain units, between a singular and a
plural. This is an essential point to which we shall revert
below, because even the best among our predecessors have
generally missed it and concentrated on formal accidents
rather than the articulation of experience. Yet even if I
disagree with the narrowly formal approach which has been
the rule in non-genetic linguistic classifications, I do not
reject the accidents of vocal forms from linguistic typology..
Nor, for that matter, should phonological typology be dis-
sociated from one based upon other aspects of linguistic
reality. Some day, linguists may raise again the question of
whether a single spectrum can be found for characterizing
a language as a whole, But we shall have to be content with
I Cf. above, pp. 24-2 5.
74
LINGUISTIC TYPOLOGY LINGUISTIC rrYPOLOGY
75
separate treatments, As Charles Bazell has aptly stated,
'structural classification must start from small systems and
not from languages as wholes, since the postulate of soli-
darity remains to be proved'. I
Phonological typology was one of the central preoccupa-
tions of the so-called Prague School in its early days. Among
the important contributions of Trubetzkoy to the Travaux du
Cerele linguistique de Prague, we find two classificatory surveys,
one for vocalic, the other for consonantal systems." It is,
indeed, symptomatic of the difficulties inherent in the estab-
lishment of a linguistic typology that, even in the neatly
circumscribed field of phonematics, we should have to set
up two distinct systems for vowels and consonants. There may
seem to exist some inverse ratio between the number of
vowels and that of consonants in a given language: in a
language with one or two vowel phonemes such as the ones
we find in the north-western Caucasus," we expect the number
of consonant phonemes to exceed 50, whereas the type of
French I use has barely 18 consonants to match its 16
vowels. Yet American Spanish, with 17 consonants, gets
along with only 5 vocalic phonemes, and Polynesian has very
short lists for both types." Consequently, languages cannot
be labelled either 'consonantal' or 'vocalic'. Besides, where-
ever we seem to discover some connexion between two
patterns of consonants and vowels, it is hardly anything
but numerical. It is only when consonants are found to make
use of distinctive features usually restricted to vowels that we
may observe some definite influence ofone system on the other,
since we usually notice that the vowel pattern ofthe language
then resorts to other features: this is true, for instance, in
a language like Russian in which many consonants are
I Proceedings of the6th International Congress of Linguists (Paris, 1949), p. 116.
2 TCLP, i (1929), pp. 39-67; iv (1931), pp. 96-116.
3 Cf. W. S. Allen, 'Structure and System of the Abaza Verbal Complex',
TPS, 1956, p. 129.
4 Cf. A. Haudricourt, 'Richesse en phonemes et richesse en locuteurs',
L'Homme, i (1961), PP' 5-10.
opposed as fronted to non-fronted and in which, in the vocalic
pattern, front is not opposed to back, but rounded to non-
lui is ([u]), but some-
times front ( [y] ), and the Iii phoneme is either front ( [i] ), or
articulated farther back ( [1] ).
This type of connexion, because of its exceptional nature,
only underlines the normal autonomy of the two phonematic
patterns. if it were shown that the efforts of some
linguists
1
to identify, on the plane ofimpressiOllistic acoustics,
the distinctive features of vowels and consonants, do not
ultimately arise from an a priori binaristic conception of
language systems, and are supported by sufficient evidence,
it would not indicate, by any means, that what is good for the
vowels is either good or bad for the consonants and vice
versa: should it be true, in one sense or another, that [a] is to
[k] what [u] is to [p] and [i] is to [t], this would not mean
that a language that presents la/ as a phoneme should either
definitely exclude Ikl from its consonantal inventory or
necessarily have it.
If, on the paradigmatic plane, little is to be gained from
trying to connect vocalic and consonantal systems, an attempt
to characterize languages by reference to the way they
balance vowels and consonants in the speech continuum
may prove more successful: as is well known, some languages,
Polynesian, for instance, never use a consonant without
tagging a vowel after it (type evcv), whereas others are not
averse to consonant clusters both before and after vowels
(cf. eeevee in English strict, eecveee in German sprichst). Some
linguistic mediums (Old Church Slavic is a good example)
present consonant clusters before vowels, but end all their
syllables in a vowel, and they could be characterized as
open-syllable languages. Modern French has a way of
restricting most of its clusters to ee by inserting a generally
non-distinctive ~ wherever it is needed to split a larger
I Cf. R. Jakobson, C. Fant, and :LVI:. Halle, Preliminaries to Speech Anafysis
(Cambridge, Mass., 1952 ) , pp. 33-34.
LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY
77
cluster of consonants ;' a phrase like Je me le demande (phone-
matically /3mldmad/, i.e, eeeeeve) is, phonetically, ~ m l g d
mad] or [3ffigldgmad], i.e, [cvccvccvc] or [ccvccvcvc].
For typological purposes it should not prove too difficult to
devise either a set of formulas of the cvev type or a short
repertory of labels: Georgian, with its frequent clusters of up
to five successive consonants.> could be dubbed a heavy-
cluster language, Modern French and Classical Hebrew
would be 'shwa' languages, and so forth. Suchlike procedures
would probably be handier and more informative than
indications on a percentage basis (e.g. e 55 per cent.-
v 45 per cent.), But, of course, languages are by no means
always homogeneous in this matter, and many of them which,
in a large majority of cases, give preference to largely vocalic
syntactic patterns, may, in a few words, indulge in very
heavy clusters: how many words like extra [/ekstra/, i.e ..
veeeev) would we have to count in French before we label it
a heavy-cluster language?
This seemingly casual but probably fairly efficient classi-
ficatory scheme suggests the use of an equally casual-looking
procedure for dealing with the specific features of phone-
matic systems. It would consist in establishing, on the basis
of our present-day phonological experience, a fairly loose
norm in reference to which the phonological system of a
given language could be characterized by indicating in what
respect it deviates from the norm, either because some of the
features of that norm are missing, or because it presents
phonological types that do not appear in it. Such a norm
should result from a survey of a large number of the most
varied languages of the world and should be set up in such
a way as to achieve the tersest possible characterizations.
The tentative one that follows is not meant to furnish much
more than an illustration of how such a system might
function.
I Cf. above, pp. I I-IS.
2 Cf. Hans Vogt, 'Esquisse d'une grammaire du georgien moderne', NTS,
ix (1938), pp. 15-16.
order to be considered normal, the phonological pat-
tern of a language should include two series of stops, affri-
cates, and fricatives opposed as voiced to voiceless, lenes to
fortes, or, in the case of stops and affricates, as non-aspirated
to aspirated. Partaking in these two series, we should expect
to find from five to ten articulatory types (orders) such as
a bilabial type, a labiodental type, and so forth. It should
be fairly immaterial whether stops and fricatives mayor may
not be conceived of as forming two parallel systems, whether,
for instance, a pair like /8/ and /0/ should be classed as
'apicals' with /t/ and /d/, or made a separate order. Unless
there were some gaps in the pattern, this should yield,
as normal, ten to twenty stops (including affricates) and
fricatives. English with eight pairs related as shown on the
following diagram
p t tJ k
b d d3 g
f 8 s J
v 0 z 3
or French with twelve stops and fricatives forming six orders
p f t s f k
b v d z 3 g
conform, in this respect, to the norm as established here.
Two to four nasal consonants, two to four 'liquids' in...
eluding vibrants, laterals, and also weakly articulated spirants
like English /r / and the weak dorsal continuant called
'Parisian r', should be considered normal.
Vocalic systems with five to ten vowel phonemes would be
considered average: five corresponds to the Spanish system
with three degrees of aperture and two series, ten to the
Danish system with four degrees of aperture and three
series (if the length distinction is disregarded), in both cases
with one phoneme only (fa/) for the widest degree of aper-
ture. The existence or status of 'semi-vowels' need probably
not enter into consideration here.
LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 79
In reference to such a norm, the phonematics of a number
of known languages could be characterized as follows:
Classical Arabic: 'emphatic' consonants, three vowels.
Finnish: a single series of consonants, gemination.
Russian: consonantal opposition between 'hard' and 'soft' ..
French: nasal vowels.
English: two sets of diphthongs.
German: affricates as a distinct series.
Italian: gemination.
Castilian Spanish: stops and non-liquid continuants
grouped in three series.
Such a norm, if widely accepted, would certainly render
some service. But we may wonder whether we have a right
to include it in a typological scheme, since it is a characteriz-
ing device pointing to deviations rather than a way of
distinguishing between a limited number of types. It is,
no doubt, likely that if our sampling were large enough, we
would find a number of languages which deviate from the
norm in just the same way: it would not be difficult to find
languages that could share with Russian the consonantal
opposition between 'hard' and 'soft'. But, still, the first step
consisting in disregarding some differences because they are
found to be very widespread, is in conflict with what is
expected from a typological approach.
A real typology would probably concentrate precisely on
what we disregarded when establishing our norm, namely
the organization of the cores' of our vocalic and consonantal
pattern. In this field, the pioneering efforts of'N, S. Trubetzkoy
must still be considered the necessary point of departure.>
I The notion of 'core system' is borrowed from a paper presented by Eugene
Dorfman at the Chicago Meeting of the Modern Language Association on
28 Dec. 1959. It designates the occlusive-fricative section of the phonematic
pattern, with whatever other consonants fit in the series and orders of that
section.
2 In YCLP, i (1929), pp. 39-67. Among recent attempts, cf. that of C. Voe-
gelin in For Roman ]akobson, pp. 598-9, and C. Voegelin and J. Yegerlehner,
'The Scope of Whole System and Subsystem Typologies', Word, xii (1956),
Trubetzkoy's main distinction among vocalic systems was
between triangular and quadrangular patterns: Spanish
with its single open vowel was said to have a triangular
pattern; Finnish, with its distinction between front lal and
back [e], a quadrangular one .. The French system was said
to be quadrangular, but it is currently losing the distinction
between front lal and back [a] and is therefore becoming
triangular. Since the physiology of speech organs implies
that there should be fewer distinctive possibilities when the
jaws are wide open, triangular patterns should be considered
normal and are actually more frequent. Well-balanced
quadrangular patterns are rare and should be dealt with as
exceptional cases rather than put on the same footing as
triangular ones. The main co-ordinates of vocalic patterns
are, of course, as was pointed out by Trubetzkoy, the degree
of aperture and the combined play of tongue and lips deter-
mining the length and shape of the front mouth cavity. The
progress of acoustic phonetic research does not imply any
revision of these basic facts. Tongue-and-lip action is excep-
tionally of a single type, as in Adyghe, more commonly of
two types (generally front-retracted versus back-rounded,
hence, for maximal oral closure Ii u/), as in Spanish or
Italian; three types when lips and tongue act independently
of each other with, as the result, Ii y ix] as in French or
German, or Ii ill n] as in Rumanian; exceptionally of four
types, viz. Ii y ill ul as in Turkish. English, with two types of
tongue-and-lip action, but with some central vowels, pre-
sents another kind of pattern. The absence of any distinc-
tion whatsoever among degrees of vocalic aperture should be
quite exceptional; two degrees are well attested (in classical
Arabic, for instance) though not frequent; three degrees are,
probably, to be considered a norm; four degrees are not rare,
but hardly very stable; since there is less space back than
pp. 444-53. Our information regarding the phonological systems of languages
quoted in what follows is generally taken from N. S. Trubetzkoy's Grundziige
der Phonologic, Prague, 1939 (= YCLP, vii).
80 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 81
front, fewer distinctions could be expected there, which is
the case in well-attested patterns such as that of seventeenth-
century French;' more than four degrees are not likely to
endure as such, neighbouring units being apt, in such
patterns, to become distinct by means of some other dif-
ference than sheer aperture..
Were it not for accidents such as two distinct phonemes
for the greatest aperture or a different number of relevant
apertures front and back, it would be easy to devise formulae
such as '23' for Spanish or Russian, '24' for Italian, '34' for
Danish, where the first digit would indicate the number of
distinct tongue-and-lip actions, and the second digit the
number of relevant apertures. Rumanian, with [in] as head
of its third tongue-and-lip type, would need some mark
distinguishing its first digit, 3, from the 3 of the Danish
formula, e.g, '323', whereas Danish would be '313', or
'3ill3' vs, '3Y3'
Many languages distinguish between long and short
vowels or between tense and lax ones, the passage from one
type of distinction to the other being often gradual, so that
both perceived duration and degree of tension contribute,
for some time, to the distinction. Should we use L for dura-
tion, and for tension, LjT for a combination of both,
we could present a formula such as '33LjT' for northern
German.
In some languages, like English, there is no sharp boun-
dary between long vowels and diphthongs; the vocalic
nucleus of beat, bait, court, and cart being treated as long or
diphthongized depending on individual or regional usage.
Here again, we could devise a formula with and D
2
for
different types of diphthongs. But I doubt whether this
would be worth while, since the English vocalic system is
quite exceptional,
Nasal vocalic phonemes are frequent only in certain sec..
I Cf. A. Martinet, 'Note sur la phonologie du francais vers 170 0 ' , BSL,
xliii (I 947), p. I 7.
tions of the world. If their pattern is parallel to that of oral
vowels, as seems to be the case in numerous African languages,
we could think of such a formula as '23N' for a pattern
made up of Ii e a 0 u I e a 0 fij; archaic Parisian French,
with IE & a 5j, would need a separate nasal formula:
'N3 2 ' .
In many American varieties of English, one would have to
take into consideration retroflex vowels, as in park, court, hurt,
which might also require a separate formula with R as the
marker. Glottalization is also to be considered a vocalic
characteristic in some languages.
What is just as important as the actual number of vocalic
phonemes is the extent to which the distinctions among them
function: in Italian, for instance, the '24' formula given above
is only true of accented vowels; elsewhere the formula is '23'
as for Spanish. In some varieties of English, it could be said
that all vocalic oppositions are blurred in totally unaccented
syllables. But, as indicated by the last formulation, it may be
preferable to deal with this type of restriction in prosodic
typology.
As regards consonants, it will be useful to distinguish
between the core system and a margin. The core system is
made up of a number of proportions whose usual diagram-
matic presentation offers a number of horizontal series, each
a succession of phonemes produced at different points of
the articulatory channel, but with a concomitant feature in
common. What I call an order is precisely the class of
phonemes articulated at the same point and with the same
organs at that point; orders appear as vertical columns on the
charts. For core consonants, we could devise formulas of
the very same type as the ones we have been suggesting for
the vowels: a first digit would indicate the number of series;
a second digit the number of orders. If we leave out the nasals
and, for Greek, the fricatives, the consonantal core of classical
Greek would be designated as '33' (jp t kj, b d gl jpht
h
khj), that
of French as '26' (jp ft s Jkj, jb v d z 3 gj).
811928 G
LINGUISTIC TYPOLOGY
Unfortunately, conflicts will arise: in English, all voiceless
consonants (the ones that have a voiced partner) could be
put in the voiceless series; but ItII and III are practically
produced at the same point and with the same organs, and
this is nearly true for ItI and 181. Therefore they can be said
to belong to the same order. This suggests that we posit two
different series for voiceless phonemes, one for stops and one
for fricatives, as shown on our diagram on p. 77. If we now
put Itl, Idl, 18/, and 101 in the same vertical row, and dis-
regard the difference between bilabial and labiodental, we
obtain, as a formula for non-nasal core phonemes, '45' in-
stead of '28' with gaps in the stop series for the sibilants and
in the fricative series for the dorsals. If, being more fastidious,
we decided that 181 is not to ItI what III is to Itfl and insisted
on making bilabials and labiodentals two distinct orders,
our pattern would be still more lacunar.
The main objection to such formulas is the fact that they
leave the prospective user in the dark regarding the nature of
the distinctions between series and orders. We might, no
doubt, use further indicators: e.g. for classical Greek, we
might write '3VA3' where V would stand for 'voice' and A
for 'aspiration', and for English '4VF5', where F would stand
for 'friction'. As regards the second digit, which indicates
the number of orders, it would be fairly easy to interpret it,
in the case of Greek, as pointing to the three most obvious
orders of labials, apicals, and dorsals, but 2 would mean 'no
labials' in Iroquois, 'no apicals' in Hawaiian, 'no dorsals' in
Tahitian. Besides, as was the case with vowels, gaps would be
left unmarked and we would never dare multiply the second
digit by the first in order to get the total number of the
phonemes concerned. We could think of giving that total
between parentheses, hence, for English, 4VF5(I6), with 16
the real total instead of 20 suggested by 4 . . . . 5. But this
would not tell us where the gaps are to be found.
We need not consider the case of the nasals and that of
marginal consonants such as 11/ and Ir I in English, French, or
LINGUISTIC TYPOLOGY
German, because it is quite obvious that no one would
bother to learn I10wto decipher formulas which would have
to be either incomplete and unreliable, or complete and
unwieldy. Yet we need not give up all attempts. According
to whether one addresses technicians or a wider linguistic
public, phonematic characterizations could be secured either
by the presentation of charts with series and orders, or by
listing the main distinctive traits and indicating the number
of phonemes affected. For classical Greek, this would give
either:
b d g dz r y ii
P
t k s e 0
ph t
h
k
h
h
E. 5
m n
it
or, less technically (and far less accurately) , 'voice and
aspiration (12 ph.), 2 nasals, 1, r; y-, and u-types,
4 apertures, length (II ph.)",
None of these solutions is really satisfactory because we
have found no way of combining brevity and precision.
There is, in fact, among the phonemes of a language much
of the same sort of solidarity as we have found among the
various planes of linguistic structure: a change here will
involve a change there and determine a whole chain of
changes. But the existence of a given type will rarely imply
the existence of another type: on being shown, for instance,
the oral vocalic pattern of a language, there is no way of
telling whether or not that language has nasal vowels as
distinctive units. Only detailed information regarding the
distribution and contextual variations of oral vowel pho-
nemes might give a clue as to whether they have nasalized
partners. A language with a 101 phoneme stands, of course,
a good chance of having a lei phoneme too, and a series of
voiced stops seems to imply a series of voiceless ones. But this
does not lead very far. Our failure to contrive a set of preg-
nant formulas need not affect us too much, however, since
LINGUISTIC TYPOLOGY
the substantial nature of the phonemes of a language has
no direct bearing upon its actual functioning as a medium
of communication. Even the absolute or relative number
of vowels and consonant phonemes is fairly immaterial in
so far as the transmission of information is concerned: in
many cases, French, for instance, will, on the phonematic
level, be satisfactorily characterized as a shwa-language with
nasal vowels.
It is not a mere chance if we have, so far, found it easier
to devise terse characterizations in reference to syntagmatic
reality than for describing the relations of phonemes in the
system: coexistence in a system means distinctiveness and
consequently as much mutual independence as is compatible
with economy; coexistence in the speech continuum implies,
it is true, preservation of each successive unit's respective
identity, but also adaptation to context, i.e, a much greater
measure of solidarity. This explains why it is comparatively
easy to devise concentrated characterizations of the syntag-
matic phonological features of a language, and why the field
of prosody is the one where typological research has yielded
the most satisfactory results so far.!
Among prosodical features, intonation is far too directly
conditioned by generally human, physiological and psycho-
logical, factors to be of any use when what it amounts to
is distinguishing among linguistic types. Only properly dis-
tinctive units, the tones, and the contrastive traits that can be
grouped under the heading of accentual prominence can be
used for typological purposes, but they do playa central role
there. In such matters, a fairly satisfactory characterization
can usually be achieved by answering, for each language, a
few questions bearing on tones and a few others concerning
accent.
I. A language has tones (distinctive tones, of course, some-
times called 'tonemes') or it has no tones: Serbo-Croatian,
I Cf. A. Martinet, Phonology as Functional Phonetics (London, 1949), pp. 1I-
15, and 'Accent et tons', Miscellanea Phonetica, vol. ii (London, 1954), pp. 13-24-
LINGUISTIC TYPOLOGY
Swedish, Vietnamese belong to the former type, English,
Russian, Arabic to the latter.
2. A tone language may present tones on all successive
intonable segments (usually syllables) or on some favoured
ones only (as a rule accented ones): Southern Chinese dia-
lects, Vietnamese, and Ibo show no speech segment without
a tone; Northern (Mandarin) Chinese, Serbo-Croatian, and
Lithuanian distinguish two or more tones, but only in syllables
with some accentual prominence.
3 All tones may be punctual, i.e. fully characterized by
their relative pitch (high, medium, or low), all of them per-
ceptibly so, or with some easily reducible exceptions (e.g,
occasional rising tones being analysed into a succession of
Jow-l-high}; or they may be melodic, i.e, opposed as rising to
falling, as in Lithuanian, as glottalized to non-glottalized,
as in Danish, as simple to complex, as in Swedish.
4 If melodic, the tones may be all on the same register as
in Mandarin Chinese or in the languages just quoted ; or they
may be on different registers as in Vietnamese, where it is not
enough to say that a tone is rising or glottalized, but where
one has to state whether it is high rising or low rising, high
glottalized or low glottalized.
5 Finally, the segment characterized by each tone may be
the syllable, as in Serbo-Croatian, a segment smaller than
the syllable, the mora, as in many languages of Central
Africa, a segment larger than the syllable, such as the (poly-
syllabic) word in Swedish or Norwegian.
I. A language may have an accent, a prominent segment
per word or comparable unit, something which is often called
'stress' in E.nglish, but accent is a functional reality, which
may involve, for its actualization, stress, pitch, length,
or any combination of these; or it has no accent: most
European languages have an accent; Vietnamese and many
Central African languages have no accent.
2. The place of the accent is either predictable and non-
distinctive, or non-predictable and consequently distinctive:
86 LINGUISTIC TYPOLOGY
LINGUISTIC TYPOLOGY
it is predictable in Czech where the accent falls on the first
syllable of the word, in Polish where it falls on the penult, in
classical Latin where its place is determined by syllable
length; it is unpredictable in Spanish where a significant
unit characterized by the phoneme succession /termino/
may mean three different things according to the place of
accent.
3- The distinctiveness of the place of accent is either
unlimited, as in Russian, or variously limited, as in Spanish
where the accent falls on one of the last three syllables of
a lexical unit.
4. If in a language tones only exist under the accent, it is
not unusual to say that that language has as many different
accents as there are tone distinctions: thus Swedish and
Lithuanian have two and Latvian three accents.
5. Most accentual languages have one accent per 'word':
in Russian, nos 'nose' loses its accent when it combines with
rog 'horn', in nosorog [nssarok] 'rhinoceros', which means
that when two lexemes are agglutinated to serve as one, the
resulting complex receives only one prominence. Other accen....
tual languages preserve the accent of each lexeme, irrespec-
tive of whether it is free or agglutinated; but some sort of
hierarchy is established among the successive accents of the
same 'word': 'rhinoceros' is Nashorn in German, and Nas-
(for Nase 'nose') and -horn preserve their respective promi-
nence, with, however, a subordination of the one of .... horn to
that of Nas-, English shares this feature with German and
extends a similar pattern to learned borrowings: energetic
with two accents as iffrom "enner-s-getic. It is true that English
has one accent in midland, just like Russian in the equivalent
sredizemnyj, and Russian has two prominent syllables in konuso-
obrdznyj like English in the equivalent cone-shaped. But,
although it is only statistically valid, the distinction between
'word'-accenting and lexeme-accenting languages may be
retained.
The preceding sketchy survey indicates how easy it would
be to develop a system of handy labels for rapid orientation
regarding the prosodical pattern of languages: German, for
instance, might be said to have an initial bound-accent on the
lexeme and no tone, if we exclude foreign elements; Italian
accent would be described as a 'word' accent, free, but nor-
mally restricted to the last three syllables. Something could
be added in both cases regarding the limitations of vocalic
distinctions in unaccented syllables, and some indications
concerning the make-up of both accented and unaccented
syllables would not be out of place.
There would no doubt remain cases where some specific
information would have to be added, if we do not want our
labelling to be misleading: to say either that French has no
accent or that it has a bound word-group accent would be
equally unsatisfactory, because the very tenuous prominence
continuing the Latin accent is one thing, the optional pro-
minence on the initial syllable is something different, and the
emphasis on and around the first consonant of the word, is
different again: the true French accent is neither impossible,
with a very slight prepausal eminence, nor impossible, with
didactic and demarcative insistence on the initial, a very
normal feature of professorial diction, nor imp-possible with
a dramatic lengthening of [p] meant to suggest anger or
passion; French has no accent in the sense English, or Russian,
or Italian has, and to use one and the same label for function-
ally so different things would be a source of lasting confusion.
Once phonematics and prosody have been taken care 04
what remains is the vast field of meaningful units and their
combinations, in other words, lexicon, morphology, and
syntax. Few linguists are likely to insist on establishing a
lexical typology, not exactly because they feel that the
lexicon of a language is too closely dependent on non-lin-
guistic reality, but-and it may amount to the same thing-
because it is what remains of the language once its neatly
structured parts have been abstracted and dealt with, in
88 LINGUISTIC TYPOLOGY
LINGUISTIC TYPOLOGY 89
other words, the domain of loosely connected units for which
it would be difficult to devise a wholesale characterization.
There is, however, one criterion whose application might
result in establishing interesting contrasts between languages.
It is what might be called the amount of motivation
1
in the
vocabulary: some languages make use of a comparatively
small stock of monemes because they frequently resort to
composition and derivation; their vocabulary may be said
to be largely motivated: such a thing being called this or that
because it is this or that. Other languages have a relatively
large number of unanalysable designations; their vocabulary
is thus more largely arbitrary in the Saussurian sense of the
term: a thing is called thus for no discoverable reason except,
perhaps, for the etymologist. The traditional illustration of
this contrast is that of German, a language with a highly
motivated vocabulary, and French, with a largely arbitrary
one: in situations where French makes use of the two mono-
monematic and highly abstract terms monte(r) /mot/ and
descend(re) /desa/ German uses combinations of two or three
monemes like aufsteigen, heraufgehen, heraufklettern, some of these
monemes being very specific and descriptive. This illustration
is further supported by the remark that French speakers are
generally satisfied with their thousands of loans from classi-
cal languages, most of which are unanalysable and arbitrary
for the vast majority who know neither Latin nor Greek.
German, on the contrary, is prone to replace foreign elements
by indigenous compounds, as when Perron becomes Bahnsteig,
Telefon is displaced by Femsprecher, and Automobil by Kraft-
uiagen.
is quite probable that this contrast between motivated
and arbitrary is something of which foreign linguists and
local purists are more keenly aware than the average users:
the present writer, a native Frenchman, had to read the
diary of Ernst Junger at the age of thirty-four before he
I Cf. F. de Saussure, Course, pp. 131-4, and Ch. Bally, Linguistique generale
et linguistique francaise, znd ed. (Berne, 1944), pp. 137-9.
realized that beaucoup de must once have been identical with
the syntagm beau coup de. It is by no means certain that
autostrade, with its unidentifiable element -strade is more
difficult to remember and to handle than its competitor
autoroute, with its obvious ingredients. If it were shown that
German children have less trouble with Fernsprecher than with
Telefon, it would probably not be on account of the lack of
motivation of the latter, but rather because Fern- and -sprech-
sound familiar, irrespective of what they mean. The French
children of today, who constantly hear the phoneme com-
binations /tele/ and /fJn/ in television and grammophone as well
as in telephone, are probably just as well off as their German
contemporaries even if they do not understand tele- and
-phone. This, however, does not mean that motivation is to
be rejected as a criterion for lexical typology, but rather that
here is one more domain where one should not jump to
conclusions.
In so far as the form of minimal lexical elements raises
specific problems, because, for instance, they are found to
vary in different surroundings, this is normally taken care
of in the morphological chapter as well as in the lexical
section of the description. This, after all, is as it should be,
since 'morphology' is the study of form, with no specification
that only the form of grammatical elements is involved. As a
matter of fact, it is more or less understood that morphology
deals only with the latter, and if the formal vagaries of
lexemes come in into the bargain, it is because they normally
result from their combination with certain grammatical
items: the shift of radicals we find in I go, I went, and the
formal variations we observe in I keep, I kept are taken care
of in all grammars of English even if lexicographers kindly
include them in their dictionaries for good measure.
Since morphology and syntax are legitimate parts of
grammar, we may now state that what remains for us to
investigate is grammatical typology. Traditional grammar
distinguishes neatly between two main chapters: on the one
go LINGUISTIC TYPOLOGY
LINGUISTIC TYPOLOGY 9
1
hand is the study of those combinations of significant ele-
ments that may involve some formal variations or accidents
(cow, cows, but ox, oxen, child, children) work, worked, but keep,
kept, sing, sang), which normally take place within the word;
this is called accidence or morphology; on the other hand,
the examination of the way separate words can be combined
into larger units, which is called syntax. But since there is
no way of defining the term 'word' in such a way as to make
the definition tally with the naive uses of it, contemporary
structuralists are prone to employ it most sparingly and to
refuse to set up any universally valid linguistic unit between
the moneme (often called 'morpheme') and the sentence.
Formal accidents will be ascribed to monemes, not to 'words',
and dealt with, for instance, as the allomorphs of a given
morpheme. They will be taken care of in the course of the
analysis of utterances into minimal significant units. Once
the identity of these units, our monemes, is established, the
linguist will have to observe 110W they combine. Since he has
already listed and described the vagaries of moneme signi-
jiants, it is by now immaterial whether the various monemes
of a given context are amalgamated, agglutinated, or form-
ally independent; the material aspect of their combination
does not concern him any more. Should one insist on positing
a unit intermediate between the moneme and the sentence, I
would propose the syntactically autonomous or independent
phrase. Such phrases would include practically all the in-
flected words of Latin or Greek, and also 'real' phrases such
as with the dog and down the road. Still, I would not be inclined
to distinguish an intraphrasal syntax, redolent of the old
morphology, and an extraphrasal one which would be syntax
properly so called.
Traditional non-genetic classification of languages was
really based on accidence; the title ofP. S. Kuznecov's short
survey, Morfologiieskaja klassifikacija jaeykoo, I is, in fact, quite
descriptive of its contents and of the way people conceived of
I In German Die morphologische Klassifikation derSprachen (Halle a.d.S,; 1956).
linguistic typology. Ifwe want to formulate what it amounted
to in terms ofmonemes, we may say that, in the last analysis,
counted was hardly anything but the degree of formal
variation of individual units, for which we can set up the
following scale:
I. A moneme may have the same signijiant throughout,
whatever the combinations it enters: the relation marked, in
English, by without is always marked by jWlloautj. This,
which logically seems the most obvious solution to the prob-
lem of the formal relations of the signijiant with its signifie,
has appeared, to generations of scholars blinded by their
admiration for classical Indo-European languages or their
ethnocentric prejudices, as a quaint and outlandish feature
when applied to the expression of functions and to gram-
matical modifiers.
2. The signifiant of a moneme may vary from one context
to another, but it will always be identifiable with a definite
segment of the utterance: Italian con, 'with', normally ap-
pears as jkonj, but also as jkoj, e.g. before the masculine
article (col, coi) and as jkolj, e.g. before the feminine article
(colla, colle); the analysis of colla into col+la is obvious,
since la is the normal form of the feminine article, and
that of col into co+l (for *con+il) cannot be said to be
arbitrary.
3. A moneme may, in some contexts, appear as a clear-cut
segment, but, in others, be merged with the signifiant of some
other moneme (or monemes), as when Fr. a, which is
generally jaj, even before the masculine article in al'hopital
(jal ... . j), is amalgamated with that same article in au
moulin (jo .. .j).
4. A moneme may never appear as a separate segment
because its signijiant is always amalgamated with that of some
other moneme (or monemes) ; still, the independent existence
of the monemes involved is never endangered: in Latin, the
dative moneme never appears as a segment which is not
at the same time the signifiant of the moneme of singular or
LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 93
that of plural; but Latin speakers were never at a loss to
distinguish dative singular and dative pluraL
5. A moneme may have a discontinuous signifiani, its use by
the speaker implying some formal modifications in two or
more different places in the utterance: the Latin maneme
usually identified as sine, 'without', was necessarily accom-
panied by a specific (,ablative') ending of the nouns whose
function was being indicated, so that, in sine dubio, its sig-
nifiant included Isine/, plus the I. . . 6j of dubio, in which,
however, the signijiant of the singular moneme was amalga-
mated. The plural moneme of French may show as a single
segment, e.g. as I. . e .. .1in dans les champs Idalefaj (cf. the
singular dans le champ Idalfaj) ; but it may also appear in the
form of several successive accidents, e.g. as I. . e . " . z .. "
o . . " mj in les petits animaux dorment Ileptizanimodormj as
opposed to le petit animal dort jlptitanimaldor/.
There is, however, another feature which has to be con-
sidered if we want to account for the pronouncements of
former typologists: that is the degree to which two monemes,
appearing in immediate succession and in a definite rela-
tionship, may be separated by the insertion of some new
element. As a matter of fact, inseparability is one of the most
useful criteria for distinguishing what is formally one word
from what is a succession of different words.' In any case, it
is the one that generations ofscribes and writers have adopted,
as a rule, throughout the centuries of alphabetic writing
practice, when they have endeavoured to divide the written
continuum of each language into those segments which con-
stitute our graphic 'words'; German spellings like aufgeben,
auJgebe, each written as one block in spite of the possibility
of inserting something between auf and the rest iaufrugeben,
ich gebe es auf), represent shocking exceptions for whoever is
not awed by the majesty of spelling traditions. In view of the
paramount role played by the notion of 'word' in former
typological speculations, it is fair to single out all the aspects
I Cf. Proceedings of the6th Intern. Congress, Paris, 1949, pp. 293-4.
of linguistic reality that afford some justification for the use
of such an ill-defined concept. Yet a clear-cut distinc-
tion between inseparability, total or partial, and variation
of signifiants is fully justified only on a strictly synchronic
plane. It is indeed clear that inseparability is conducive to
formal accidents which ultimately result in amalgamation:
phonetic evolution will tend to merge successive elements in
the utterance, and this can be counteracted only inasmuch
as the elements in question may, at any time, cease to be in
contact; any limitation of the latitude to combine monemes
freely is the first step toward eventual coalescence, semantic
as well as formal. WIlen what is at stake is a general charac-
terization of languages, from a synchronic angle, no doubt,
but on a wide temporal scope, widespread inseparability will
be found to have bred some degree of formal overlapping
and amalgamation, so that the two features normally go
together.
Sapir's outstanding merit was to discover .and point out
that the traditional approach to non-genetic classification
was founded upon a fairly casual rating of the frequency and
degree offormal adhesion, overlapping, and coalescence. He
therefore recommended a typology that would not be based
exclusively upon features which might not be so fundamental
as many scholars had believed them to be. As a matter of
fact, observation had revealed that most languages presented.
a mixture of various formal types, and comparative evidence
had indicated that formal features were subject, through
time, to considerable variations either towards syntagmatic
coalescence or away from it. Without rejecting what he
called degrees of 'fusion' and the traditional 'synthetic' I"'J
'analytic' opposition, Sapir concentrated on 'the nature of
the concepts expressed by the language' and made it the real
foundation of his classificatory attempt. He thereby un-
doubtedly went farther and deeper than any of his predeces-
sors. A proof of this is that, to this day, forty years after the
publication of Sapir's book, anyone who deals with language
94 LINGUISTIC TYPOLO G Y
LINGCUISTIC TYPOLOGY
95
classification necessarily uses Sapir's analysis either as a
starting-point or as a frame of reference. Yet, for all its
perspicuity, it stands as a nearly tragic illustration of the pit-
falls of psychologism.
Sapir's original contribution centres around a conceptual
analysis of language elements starting from the acceptable
view that we should find, in all languages, significant units
that do not carry in themselves a mark of their relation to
the rest of the utterance, say, chair, lamp, and others, such as
with or for, that are expressly meant to indicate what sort of
relations the former keep with each other. This leaves out
the theoretical possibility of a language entirely made up
of words of the yesterday type, which stand, at the same time,
for some element of experience: 'the day before this day',
and the relation of that element to the rest of the experience:
'as the segment of time when ...'. But we know that economy
rules out any such language, and we may agree with Sapir
that all known languages utilize monemes with the chair type
of function, and monemes having the with type of function.
Our criterion for distinguishing between one type and the
other is not semantic, however, but positional: if chair by
itself is to have a function in an utterance, it must stand in a
well-determined position in relation to the other elements of
the utterance; but this changes as soon as a moneme of the
with type is added to it: the phrase with (the) chair will not
depend on its relative position for the expression of its
function. The importance of this reliance on formal criteria
will appear in what follows.
As a third type of element, Sapir poses what he calls
'derivational concepts', which formally correspond to affixes.
There can be little objection to this, except that, of course,
his conceptual approach prevents him from connecting com-
position and derivation, which are such closely related
phenomena that the analyst is often at a loss to say, in con-
crete cases, whether a given moneme is an affix or an element
of a compound word: in contemporary French, some speakers
use the form bus as an abbreviation of autobus; this form was
probably used first in imitation of its English equivalent, and
it is not in general use yet; for those use autobus is
a compound word of the autoroute 'motor-road' type; for those
who do not, it is a derivative of the electrobus type, with bus
as a suffix. The formal criterion of both derivation and
composition is, of course, a combinatory comportment of
derivatives and compounds which is, in all respects, identical
with that of corresponding single monemes, It is, no doubt,
quite essential to determine whether a language makes use of
derivation, or composition, or both: languages vary a good
deal in that respect. But this has little to do with the distinc-
tion between monemes that do not indicate their relations
to the context (dependents) and those that are meant to
mark those relations (functionals); both composition and
derivation yield units which may function as dependents
(farm-yard, yellowish) or as functionals (on-to, Lat. in-ter,
in-tr-a, in-tT-v) . We have here a distinct type of linguistic
activity, which we could dub lexical, or, more generally,
paradigmatic expansion. Any typology will have to take this
into account in connexion with the lexicon, a domain in
which, as we have seen, the frequency of still analysable com-
pounds and derivatives could be characterized in terms of
motivation.
Sapir further distinguishes between 'pure relational con-
cepts' of the with type, and 'concrete relational concepts', but
he never makes perfectly clear where the difference between
the two types actually lies. It would seem that this distinc-
tion more or less parallels that between what we call func-
tionals and modifiers, the grammatical elements which secure
autonomy, and those which do not. But his conceptual
approach misled him into expressing the difference in terms
of degree of abstraction, and made him fall a victim to the
common illusion that some grammatical words or items (his
'pure relational concepts') have no meaning. If meaning is
to be mentioned in linguistics at all, it should be defined as
resulting from the necessity, for the speaker, of choosing at a
point among several units for the expression of some element
of experience.. Choice, on that level, implies meaning, and
meaning is there only if there has been a choice. Syntactic
functions, such as the ones expressed by prepositions or
cases, are, in a way, predetermined: a dative relation is
expected after a verb meaning 'to give'. But the choice often
exists between the presence of the dative complement and its
absence: advice can be given absolutely or specifically to
someone. A grammatical subject, as such, is an item whose
presence does not result from a choice: the speaker does not
choose to use a subject or not, because, by definition, as it
were, a subject is what must be added to a predicate to make
a statement in those languages where we have a right to
speak of a subject. Therefore, since the introduction of the
subject function is never a matter of choice, we may say it
is deprived of meaning. A formal consequence of this is
a tendency to eliminate any mark of the subject function, so
that a zero signifiant should correspond to a zero signifi. But
this is an extreme case. The difference between functionals
and modifiers, i.e, the two types of grammatical monemes,
should not be expressed in semantic terms, although modi-
fiers are likely to be semantically richer than functionals. It
should be based upon the part each type plays in the sen-
tence. The notion of relation, which Sapir used in order to
give a semblance ofunity to his non-lexical units, is extremely
vague and misleading: there is no doubt that a definite
article normally involves some relation to what precedes,
since one of its normal functions is to indicate that the
accompanying being, object, or notion has been previously
mentioned. But this role is in no way distinct from that of
pointing to the fact that the being, object, or notion is uni-
versally known, as when we say the sun, the King, or the lion
for the species. The definite article is not indicative of any
relationship; it is the mark of some specific information
added to the communication, just like any other modifier such
9
6 LINGUISTIC TYPOLOGY LINGUISTIC TYPOLOGY 97
as 'singular' or 'plural'. It summarizes some previous informa-
tion, but gives no indication as to what the function of its
environment is the linguistic rendering of experience
which is being communicated. Amalgams may make it
difficult, in many cases, to decide which phonic segment is
to be ascribed to a functional, and which to a modifier.. But,
functionally, the two types are fundamentally different.. The
distinction is not, as Sapir has it, between material content
and relation, but between functional monemes and non-
functional ones, the latter including the bulk of what Sapir
puts under 'material content', namely lexicon, including
elements of composition and derivation, together with our
modifiers.
It is, as we have seen, by no means impossible to find
functional justifications for everyone of Sapir's four con-
ceptual groups. But this does not mean that we can retain
Sapir's analysis as the foundation of a typology that tran-
scends formal accidents. Once we have ruled out reference
to things meant as a principle of linguistic classification, it
becomes clear that the four types of facts are not comparable
and that it makes no sense to consider them along a scale
leading from pure conceptual to pure relational, Should we
try to classify monemes according to their function, we
should first put on a plane of its own the distinction between
derivational monemes and the rest. The latter could not very
well be dubbed non-derivational, since the fact of being
derivational implies nothing but a limitation, in the distribu-
tion, to cases of paradigmatic expansion, in other words,
a negative feature. We certainly would not want to have
to specify 'non-derivational' every time we characterize
a moneme affected by no such limitation. Once the problem
of derivationals is set aside, our two criteria would be
functional indication and grammaticality. By combining
them, we would obtain four types again, as shown on the
diagram on P: 98, out of which only three (viz. I, 3, and 4)
could be found parallel to those of Sapir.
811928 H
9
8 LINGlJISTIC TYPOLOGY LINGUISTIC TYPOLOGY
99
Two of them (I and 2) would not partake in functional
indication or, in other words, would not confer, to themselves
or others what we have called syntactic autonomy: the
ones (I) would be lexical dependents, the others (2) t ~ r
grammatical. dependents such as pronouns, or grammatIcal
determinants, i.e, modifiers. The remaining two would be
autonomous monemes of theyesterday type (3) and autonomy
conferring grammatical items, i.e, functionals (4). Predica-
tives, as such, would not all belong to the same one of the two
lexical types, since they might be said to be dependent where
they can be orientated (in languages with both active and
passive voices), but autonomous, even independent elsewhere.
Furthermore, there is a fundamental difference between
Sapir's groups and our types, which makes it impossible for
us to utilize the precedent. When we say that our types are
functional, we imply that a given item may belong to dif-
ferent types: in a large number of languages, the same
moneme is used predicatively as the equivalent of 'to give'
and as a functional with the value of 'dative'; this; of course,
does not entail that all predicatives can be used as func-
tionals or all functionals as predicatives. In Basque, the
moneme -ko is used either as a functional ietxe-ko 'of the
house') or as a derivational use for lexical expansion (etxeko-
a-k 'the domestics'; -a-k = 'the' -l-plural}; here again, this
does not mean that all functionals of Basque can be used as
derivationals, but it indicates that our types are not mutually
exclusive. We should not operate as if a given language had
or had not modifiers or functionals, but assume that any
lexical
without indication of function
chair
3
lexical
function indicating
yesterday type
2
grarnmatical
without indication of function
the
4
grammatical
function indicating
with type
language is likely to have monemes acting as
Therefore we cannot follow Sapir when he classifies languages
according to whether they 'express concepts' of this or that
group of his, which, in less-guarded parlance, would have
been expressed as 'having' this or that sort of units: beside
his group I (of 'basic concepts'), which he sensibly assumes
to be universal, a given language should use, for the ex-
pression of relations, either his group (of 'pure relational
concepts') or his group III (of 'concrete relational concepts') ;
of his four, A, B, C, D, classes, two (A and B) combine I and
IV, and two (C and D) combine I and III; the absence of
derivation distinguishes A from Band C from D. Groups IV
and III are thus presented as mutually exclusive, which rules
out any attempt to preserve some parallelism between our
types and Sapir's groups: when Sapir conceived of his group
III, he must have had in mind concepts corresponding to our
modifiers; when he set up his group IV, he certainly wanted
to include in it elements corresponding to prepositions, which
are the most obvious of our functionals, and position, as a
mark of function, is expressely mentioned in this connexion;
but the lack of perspicuity inherent in the conceptual ap-
proach led him to list English and French among languages
that make no use of his group IV, thereby disregarding the
fact that in both languages the relations of nominaIs with the
rest of the utterance are, in the overwhelming majority of
cases, indicated by means of position or prepositions. It
would be interesting to look for the features of these two
languages, such as widespread amalgamation, concord, and
discontinuous monemes of all sorts, that must have been
instrumental in obscuring for him such an obvious fact. But
this would lead us too far. It will suffice to point out here
that, even in his conceptual classification, Sapir is so depen-
dent on form and intent upon finding formal differences
supporting his conceptual grouping, that he indicates, as the
only normal expression of his 'concrete relational concepts'
(group III), affixation and inner modification, the use of
concepts' (group IV).
As soon as each utterance in any language is conceived
of as a succession of minimal signs, with or without over-
lappings, it becomes clear that grammar can be presented
as the set of rules restricting the free combinations of all these
signs. A grammatical typology is bound to start from the
various types of restrictions that have been, inductively or
deductively, identified. We have seen, in a preceding chapter, I
that a language can hardly be conceived of without such
restrictions, for instance, regarding the extent to which the
various monemes may assume the various existing functions.
Some monemes must necessarily be identified as endowed
with functions and others as functional indicators. This
accounts for the fact that there is no language recorded in
which one should not somehow distinguish between gram-
matical and lexical items, whether these be opposed as empty
words to full words, or as grammatical categories to basic
concrete units. Restrictions on the free use of the different
lexemes in all the existing functions are probably to be found
everywhere, but in many cases these may result from the
specific meaning of some lexemes: a thing may not be used
as an agent; some predicates cannot be passively orientated.
But what may be peculiar to some languages and therefore
typologically relevant are wholesale restrictions resulting in
splitting the lexical repertory into neatly contrasting classes.
In this respect the most fundamental dichotomy is the
one between languages in which all lexical monemes can be
found performing the same basic functions, and those, prob-
ably a clear majority in the world of today, in which some
monemes, which we may designate as 'verbal', are specialized
and restricted to predicative uses: on the one hand, we have
languages in which the equivalent of 'tree' and 'it stands
upright', of 'leg' and 'it walks' are identical and distin-
guished only by the predicative use of the latter, some non-
I Cf. above, pp. 63-64.
predicative function the on the other hand, forms
of speech in which the equivalents of 'tree' or 'leg' may per-
haps be to act as predicates, whereas those of 'it
stands' and 'it walks' can be predicate and nothing else.
Languages with a special 'verbal' class restricted to pre-
dicative use can be further subdivided into languages in
which non-verbal lexemes can normally be used with pre..
dicative function and those in which this is formally im-
possible. Russian is a language in which lexemes endowed
with 'nominal' functions are also found as predicates. On the
contrary, English is a language in which 'verbal' monemes
have a practical monopoly of predicative function, and it
illustrates a linguistic type where a given lexical item is
always used either predicatively (verb) or non-predicatively
(noun or adjective); in regular, out-of-situation, syntax, no
statement is possible without some verbal form.
Another fundamental distinction should be made between
languages in which the predicative moneme, once it has
been selected, can only stand in one definite relation to every
one of the other elements of the utterance, and languages in
which the predicative moneme can be orientated in different
ways in order to give prominence to its relationship with one
of the other elements. This means, in traditional parlance,
that great importance should be attached to whether a
language distinguishes or not, in what we may call its con-
jugation, between an active voice and one or more so-called
passive voices. This feature may be connected with the
presence of a subject properly so called, i.e. the compulsory
use of an item accompanying the predicate and whose rela-
tion to it is accordingly established as preferential and
fundamental,
All this amounts to saying that the foundations of a lin-
guistic typology are to be looked for in the way in which each
language community proceeds to analyse experience into a
number of elements in such a way that the linear order of
their succession will not prevent hearers from perceiving the
100 LINGUISTIC TYPOLOGY
'Y'pC"p'Y''trpri to that of 'pure relational
LINGUISTIC 'TYPOLOGY 101
102 LINGUISTIC TYPOLOGY
nature of their mutual relations and thus reconstructing the
total experience. The basic syntactic patterning of each
language has to be presented in terms of minimal significant
units, the monemes, each moneme being conceived as the
habit of identifying some definite facet of experience with
some formal distinction. There are thus always two faces
to every moneme, It would be linguistically ideal to have
a situation where the formal face of every moneme would
be always the same and correspond to a neatly analysable
segment. In fact, the signifiant of a given moneme may vary
according to context, be discontinuous, or Inerge with neigh-
bours, so that only functional identity salvages the unity of
the moneme, A fundamentally synchronic description of a
language should go beyond these formal accidents, and reach
the essentials: the resources of a language as an instrument of
communication.
I
LINGUIS to IE
I
N this world of ours where, apart from pitiful exiles, every
man is supposed to owe allegiance to some ruler or State,
this allegiance is tacitly assumed to imply the use of one
and the same language: an Englishman speaks English and
a Frenchman speaks French. Belgians are widely believed to
speak 'Belgian', a belief that is partially substantiated by the
existence of Flemish, It took two world wars to convince the
French that the British and the Americans were different
nations, but the realization of their separateness carried along
with it the by now well-entrenched belief in the existence
of a fully distinct American language. We all, in daily life,
speak, and sometimes act, as if there existed neatly circum-
scribed language communities wherein all members are
expected to behave linguistically in exactly the same way.
Those who do not, in all details, are said to speak with an
'accent' if their deviations from an assumed norm are mainly
phonic. They are supposed to speak a 'dialect' if their aber-
rance extends to grammar and lexicon, particularly if com-
munication is thereby somewhat impaired.
As long as linguists were 'philologists' who mainly concen-
trated on written, literary forms of human communication,
they were not inclined to take exception to this sketchy and
naive approach to socio-linguistics: 'accents' were hardly
ever noticeable as such in their texts, and 'dialects' were
best forgotten about except, perhaps, in the isolated case of
ancient Greek..
But even after they have become conscious of the basically
vocal form of language, after decades of efforts devoted by
many of them to the collection of dialectal material, they still
seem to take it for granted that there are well-identified
14 LINGUISTIC VARIETY LINGUISTIC VARIETY
15
objects called 'language communities' whose members speak
alike in all respects. At least most of them behave and write
as if they did. In this matter there is little difference between
'traditionalists' and 'structuralists': all would seem to believe
in the perfect unity of each language. Among the latter, few
would hesitate to posit that French has so many (e.g, 34)
phonemes, or that the /s/ of Spanish is apico-alveolar, even
if they were ready to grant, when challenged, that many
French speakers use less, and many others more, than thirty-
four phonemes, and that millions of Spanish speakers, in
Spain and elsewhere, use dorsa-alveolar sibilants.
The homogeneity of linguistic communities is a useful
pragmatic assumption, at least at a certain stage of linguistic
research.. If languages are, first and foremost, instruments
of communication, it is fairly natural that we should assume,
at least as an ideal, that all people who use one of them share
the bundles of articulatory habits and vocal reactions to
various stimuli whose sum total we call a language: com-
munication would be best secured if all people concerned
spoke exactly in the same way. Some variety is no doubt
welcome in human affairs. But when efficiency is at stake,
relevancy is what counts exclusively. But unbiased observa-
tion shows that when people understand each other they do
so in spite of differences which affect not only irrelevant
trimmings, but sometimes fundamentals. It is true that
when we look at things from a dynamic angle, the prospect
clears up a little: it is comforting to notice that when people
actually communicate they tend to identify their speech
habits and linguistic reactions: communicating improves
communication and does it apace. But this might eventually
result in the complete unification of linguistic communities
only ifit worked within their confines exclusively, which is by
no means the case. Linguistic convergence is universal: just
like charity, it begins at home and, just like charity, it extends
to the whole of mankind; it takes place among those who
feel they belong to the same language and social group and
believe alike in all respects, as between a new-
comer and the former residents a suburban district; but
it takes among Russian and
fishermen who to settle in the same neighbourhood
along the and results in the development of
a new form of speech..
1
convergence will inevitably
breed divergence: the new-comer who adapts his speech to
that prevailing in his new will thereby deviate
from what had been his set of linguistic habits so far, and all
the quicker if the original linguistic differences between the
two parties did not hamper immediate oral communication..
As long as needs differ from one district to
another, complete linguistic homogeneity cannot exist.
Ifit has so for such an obvious fact to be acknow-
ledged, it is again because those who were interested in
matters of language concentrated on literary forms of com-
munication.. Those forms had early assumed a large measure
of unification since they were used by somewhat restricted
groups of literate people engaged in a sort of communication
which did not require the establishment of physical contact
and which, consequently, could easily spread to the most
distant provinces..
Once the familiar and comfortable idea of the homo-
geneity of linguistic communities is abandoned, the world
appears as an ocean of conflicting attractions, convergence
here breeding divergence there, with new centres of attrac-
tion developing at all times and threatening to disrupt
existing ensembles. this is true on all levels, national,
provincial, local, and familial. In the practice of descrip-
tive linguists the of this endless linguistic
variety has led to the position that if a linguistic description
has to be .consistent, it must be that of an idiolect, i.e. the
language as spoken by a single individual. But does this not
contradict our assumption that language is above all an
I See Olaf Brach, 'Russenorsk', Archivfur slavische Philologie, xli (1927), pp.
29-62.
106 LINGUISTIC VARIETY LINGUISTIC VARIETY 17
instrument of communication? Besides, if we want to do
justice to all the implications of the ceaseless turmoil we have
just sketched, we cannot be satisfied with limiting our obser-
vation to one single individual, since that individual will not
handle his language tomorrow in exactly the same way as he
does today or did the day before. The idiolect, as a frame
of linguistic description, needs to be precisely dated.
At this stage, and before we proceed, it may be useful to
stress that the rather startling picture that has just been
presented is no fiction, but that it is derived from factual ob-
servation: when, in 1941, sixty-six French officers, born and
reared in Paris, were asked some forty-odd simple questions,
such as 'Do you pronounce patte and pate alike?', which were
meant to reveal the main features of their respective vocalic
system, no two of them agreed in all respects, although their
phonological comportment as a whole contrasted with that
of their 343 non-Parisian comrades who also answered the
questionnaire. I
As regards the evolution an idiolect may undergo through
time, the present writer may be allowed to refer to a sketch
of the system of French vowels he published in 1933.
2
In that
first attempt to present a phonological analysis of that
language, being still unaware of the existing latitude in such
matters, he identified his then system with that of French at
large, and posited two long phonemes jy:j and jre:j as dis-
tinct from jyj and [ex] on account of such minimal pairs as
sur-sure (/syr1"'-' Isy:rI), seul-seule (/srelj"-' Isre :1/). When recon-
sidering his phonological practice, some years later, he
noticed that these distinctions of length had been wiped out,
probably under the influence of his Parisian surroundings.
If such variations from speaker to speaker and, with the
same speaker, from one period to another, are attested in the
realm of phonology, a domain where people as a rule are
I Cf. Prononciation generally, and Phonology as Functional Phonetics, pp. 36-37.
2 'Remarques sur Ie systeme phonoIogique du francais', BSL, xxxiv (1933),
pp. 191- 2 0 2
least conscious of what they do, they can hardly be doubted
in the fields of syntax and lexicon. The practical conclusion
to be derived from this is that a linguistic description that
is not expressly that of a dated idiolect should be expected
to subsume divergent usages. If existing variations are not
sacrificed on the altar of descriptive simplicity and are duly
presented, the result will be the establishment of some sort of
hierarchy among linguistic oppositions: some distinctions are
found to be universal among the members of the group under
consideration; others are found to be kept by some members
only and to be disregarded by other members. Ifthe language
at stake is American English as a whole, it will be pointed out
that horse and hoarse, morning and mourning are kept distinct by
some speakers, while others pronounce them alike, and this
type of distinction will not be put on the same level as that
between card and cord, lardand lord which seems to be univer-
sal. In describing French in general, it cannot be said that
pres differs from presin the same way as pris differs from pres,
although physically the three vowels are close to cardinal [i e ],
because all French speakers distinguish pres from pris, but
millions identify pres and pres.
We can conclude from all this that the notion of linguistic
community is not only useful, but unavoidable in linguistics
as soon as a language is conceived as an instrument of com-
munication constantly adapting itself to the needs of the
group who make use of it; 'communication' implies 'com-
munity'. But in order not to let this term confuse linguistic
issues, it is indispensable to stress a number ofwell-established
facts.
I. No community is linguistically homogeneous. no two
persons use a language in exactly the same way; the same
situation will elicit different linguistic reactions from different
onlookers; no two persons will use or understand the very
same vocabulary; even the highly structured aspects of
language, such as phonology and morphology, may differ
in important matters from one speaker to another without
impairing mutual understanding and even without being
noticed by the interlocutors.
2. Many people belong to two, or more than two, com-
munities. This is, of course, the case in such well-known
bilingual stretches as Brussels, Alsace, or South Africa. But
this applies to many situations where both a vernacular and
a standard language are alternately used by the same people
with different interlocutors, From a linguistic standpoint
we cannot make bilingualism depend on the amount of
prestige enjoyed by the two forms of speech in contact. The
alternate use of two different phonological systems is probably
the least ambiguous test of a bilingual situation.
3 Many people use concurrently different styles of the
same language. The same Frenchman may use from one
minute to the next two totally different equivalents of 'shall
we go?'; the literate partons-nous? or the slangy on les met?
with the familiar plural first person pronoun on.
4 Many people who do not use more than one style or one
language understand different styles or different languages.
Passive knowledge of languages and unimpeded aural under-
standing of various not actively used styles is of frequent
occurrence and should playa great role in the correct appre-
ciation of socio-linguistic situations..
Before we examine in detail cases where people cannot help
being aware of linguistic differences, it is important to revert
to definitely unilingual situations where the different mem-
bers of the community are not aware of any such deviation
in the speech of others or in their own as could be dubbed
'accent' or 'dialect', In these the communicative ideal of
language seems fully realized: it really does not matter in the
least whether the speech of one man is physically identical
with that of another provided possible divergences do not
interfere with the free and easy transmission of'experience. In
this respect it is quite essential that all participants should be
intimatelyconvinced that theyspeak 'the same language': since
language is constantly redundant, words having generally
more phonemes than is strictly necessary to keep them
distinct from others in given contexts, and messages being
as a rule more profuse than logic a few
divergences will not prevent nor even impair mutual under-
standing. People being normally intent upon understanding
what is said, and hardly upon observing the way it is said,
they will never notice any such divergence. Language works
best when it is not noticed as such, and speakers will be better
off as long as they manage to forget about it. This is why they
are apt to get impatient and scornful when some 'accent' or
'dialect' feature reminds them that linguistic communica-
tion can be a problem. Unexpected deviations may prove
misleading even if they do not affect the distinctive and
significant pattern: if I have never heard my la/phoneme
pronounced as anything but [a], an occasional [0] rendering
may be startling and may keep me wondering for a split
second whether that unexpected sound 'meant' something,
even if the nearest vowel in the system is some distant /'JI
phoneme.. But if one is used to hearing something that
deviates from one's own practice, that something will cer-
tainly pass unnoticed: if I pronounce my /a/phoneme as [a]
but am quite used to hearing [0] for the same unit, no [0]
pronunciation can ever ruffle me except, perhaps, if it
comes from someone from whom I do not expect anything
but [a].
In this connexion I may perhaps be allowed to present an
illustration borrowed from my own experience: as a child I
knew that my mother's French was slightly tainted by her
early use of a Franco-Provencal dialect and that, on the other
hand, when she spoke that dialect there was something that
did not sound quite right. One day, about the age of twenty,
I was sitting in an orchard reading for an examination. My
mother was conversing with her sister, about ten yards away.
At some point she happened to mention the name of the
famous political leader Marat and pronounced it [mana]
with a dorsal spirant. This startled me out ofmy reading and,
108
LINGUISTIC VARIETY
LINGUISTIC VARIETY log
These look very different. But it happens that contacts be-
tween people with such divergent vocalic systems have been
so frequent during, say, the last two centuries, that the
French language is handled in such a way to render the con-
fusion of lal and [a], lei and [e], 101 and Idl (as sur ce ==
sur ceux) perfectly innocuous. Only someone intent upon
detecting phonological differences would think of my wife's
speech and my own as two distinct varieties. The only valid
criterion in such matters is less perhaps the ease of under-
standing, which we would not know how to measure, than
the existence of bilingualism, in the widest sense of the word,
with in terms of dialects. Other terms such as patois, brogue,
bable, Platt, refer to varieties which are ultimately presented
under the rubric of dialectology. attempt on the part of
specialists to use these terms without previous redefinition
is likely to be unsatisfactory because they are, in fact, quite
unspecific designations used, with a touch of contempt, by
standard speakers in reference to any speech variety that is
not socially acceptable. The American College Dictionary defines
patois as 'any peasant or provincial form of speech', which
amounts to authorizing the use of that term for any language
variety except some metropolitan standard. Such a defini-
tion, which reflects social prejudices rather than the realiza-
tion of real linguistic differences, makes the word useless for
our purposes unless we decide to define it in our own terms.
When we try to approach this problem from a linguistic
angle, it is disappointing to find that little can be done if we
start from those achievements of which we are so proud, viz.
our phonological and morphological patterns. Here is my
wife's vocalic pattern in French in absolute final position pre-
sented side by side with mine:
LINGUIS1'IC VARIETY
110 LINGUISTIC VARIETY
halfjokingly, I told my mother she had no right to pronounce
that name otherwise than [mara], i.e, with a trill. It suddenly
had dawned upon me that she used both 'rolled' and 'non-
rolled' r, which explained why her French sounded provincial
and her dialect Frenchified, but that she usually had the trill
for single intervocalic r of the spelling and the dorsal spirant
for -rr-. Her use of the latter in Marat, probably as an imita-
tion of the teacher from whom she had first heard the name,
had suddenly made me realize that I expected from her
a tongue tip [r] wherever the spelling had a single r.
Generally speaking everything one has grown up with
belongs to one's universe, and this applies to linguistic ex-
perience as well as to any other domain. Anything we have
heard in the course of the acquisition of our language is felt
to be part of that language. This does not, by any means,
imply that we ourselves shall make use of everything we
readily accept from others: a number of words, certain pro-
nunciations, we have heard from people we either do not
approve of or dare not imitate, and these we avoid. But this
remains unconscious until, by chance, we are startled by
hearing them in an unexpected situation or context.
Something similar applies to different generations in
each other's company: we hear in the speech of older people
many turns and traits which we would never use, but which
still belong to our language. Yet our grandchildren will
never get used to them and these features will thus gradually
disappear, first only from active use as long as we live, then
also as acceptable forms as soon as only those generations
remain on earth that have never heard them. What we have
heard ever since we could perceive what was said around us
is always part of our language as we conceive it, even if
we may exceptionally become aware of the obsolescence of
certain terms or elements.
Linguistic variety on a large scale, within one and the same
community, is, by both laymen and linguists, usually dealt
1
e
y
o
a
u
o
1
e
e
a
y
e
u
o
I I I
112
LINGUISTIC VARIETY LINGUISTIC VARIETY 113
namely, the use same persons of two phono-
logical and morphological patterns depending on one's inter-
locutors.. need a say, ' to refer
to linguistic forms used by unilinguals their oral com-
munications with any other member community, even
those who use some other dialectj : the York City form
of speech and the Chicago form of speech would thus be
labelled two dialects, of American since a speaker of
one would not hesitate to use his own form of speech when
addressing a speaker of the other, term, say 'dialect,',
desi?,nate linguistic forms used as a vernacular by
bIlInguals In their communications with some .particular
members of the community, whereas they use a dialect. with
the other members; the speakers of any dialect are in fact
2' ,
a smaller (provincial) community within the larger (national)
community.. This type ofsituation will be amply illustrated in
what follows ..
It is clear that dialects., are likely to be more divergent than
since if they were not and i:f, accordingly, their use
did not hamper mutual understanding, speakers of dialects,
would not take the trouble of learning a dialect which is not
h
. 1
t eir vernacular.. this greater divergence is not included
in the definition because it is, as we have seen, too difficult to
measure..
Whether dialects. all enjoy the same prestige, or whether
one of them, or a group of them, ranks higher and is, there-
fore, rIO longer considered a dialect, but the standard, is
again a different matter: prestige is difficult to measure and
better not intervene in linguistic classifications.. Still, it is
Important to be aware ofsuch as situation in the
United States, where no regional standard can be localized
. '
and that In Italy, where the Tuscan varieties when de-
. ,
prived of their strictly local features, rank definitely higher
than other dialects., which are; of course, and in contra-
to dialects., local forms of the national language.
It IS worth remembering that the original use of 'dialect' in
ancient Greece, before the establishment of Athens' cultural
hegemony, corresponds to our and reflects a
linguistic situation far more similar to the American than to
the Italian one.
It could be objected that if dialect, necessarily implies
bilingualism in contradistinction to dialect., it would be
better and clearer to call dialect, a language, a term which is
implied in 'bilingual'. But a language is understood to enjoy
a status which can by no means be granted to many dialects,
that only survive as the impoverished mediums of retarded
rural segments of a community. The widespread reluctance
to speak of bilingualism in the case of situations involving
dialects, is due precisely to the impression that they are
granted thereby a status they do not deserve.
The dialectal situation in France has long been the subject
of careful and detailed studies, and although professional
dialectologists have widely disregarded some of the distinc-
tions which we today find basic, these studies have largely
contributed to making France one of the best fields if we
want to illustrate linguistic variety within one and the same
community.
It cannot be said that France, as a political entity, coin-
cides with a linguistic community, since millions ofunilingual
French speakers live in other countries such as Belgium,
Switzerland, and Canada. But it cannot be doubted that any-
where in France people are supposed to be able to handle
their public and private affairs in French: all children
living in France are expected to attend schools where
French is taught as a subject and at the same time is the
medium through which other subjects are taught; further-
more, all able-bodied males spend an average of two years
in an army in which French is the only official medium.
As a result of this, non-French-speaking Frenchmen must be
rare except among older people, chiefly women and mainly
in Alsace and the traditionally Frankish-speaking fringe of
Lorraine.
811928
114 LINGUISTIC VARIETY
LINGUISTIC VARIETY 115
At the present day a definite majority of the 45 million
of must unilingual French speakers,
which, of course, does not mean that they all speak alike. All
of them could be said to make use of dialects., i.e. different
varieties of French, but this would by no means reflect the
way the French react to such variations: deviations from
what is felt to be the norm in matters of pronunciation are
labelled 'accents'; other aberrances, if startling, might be
considered 'patois' by those who would be tempted to apply
this term to any departure from the norm. The few competent
people who have dealt with French dialects! refer to them as
francaislocaux. Some of themhave been summarily described. I
They are certainly well attested in Gillieron's Atlas.' but
there is no way of knowing with certainty whether what we
find at a certain point is dialect! or dialects, In a radius
of sixty miles around Paris, Gillieron's forms are likely to
originate from unilinguals and therefore to represent local
French (dialectsj), In the southern half of the country all
notations, with very few exceptions, should represent utter-
ances from bilinguals who knew French at least well enough
to understand Edmont's questions and who translated the
French words into the local dialect.; But, in between, the
forms attested in many places are of such a nature as to
authorize anyone of the two interpretations.
Unilingual France is expanding rapidly. The First World
War was fatal to dialects. in many sections of the country:
all fit men between the ages ofeighteen and fifty went to war.
At first they were in locally mustered regiments, but very
soon, heavy losses resulted in successive amalgamations, as a
result of which soldiers from all parts of the country could be
found in the same unit. This meant that there was little use
for dialects, at the Front and behind it, except when such a
I e.g. Le Francais parle a Toulouse, by Jean Seguy (Toulouse, 1950) (Biblio-
theque Meridionale publiee sous les auspices de la Faculte des lettres de Toulouse,
rst series, XXVIII), 132 pp.
2 About the Atlas linguistique de la France by J. Gillieron and E. Edmont, see
Sever Pop, La Dialectologie, I. Dialectologie romane (Louvain, n.d.), pp. I 13-36.
one was still practised and understood over a large area so
that there might remain, in some regiments, groups of
soldiers who found it easier to use it among themselves than
French. On returning home, those who for close on four
years had been using French almost exclusively insisted on
speaking French to their children, and their wives eventually
followed suit. This explains why, in large sections of present-
day rural France, dialects, are frequently understood but
hardly spoken by people under forty. It is likely that the
Second World War has sealed the fate of a good many of
those which had fully survived the First. The very possibility
of a dialect disappearing through a break in the transmission
is a clear indication that that dialect is a dialects. A dialect!
cannot disappear, since, in the case of unilinguals, it is the
only form of speech at their disposal. The effect of a war will
be to make it closer, if not to the 'standard', at least to the
average form of speech.
It is a great pity that Gillieron was not aware of the neces-
sity of distinguishing between the two types of dialect and
consequently did not try to devise a way of drawing the
frontier between unilingual and bilingual France in his time.
It is true that the thing would not have been as simple as
we have made it sound: there are socio-linguistic situations
where the shift from one form of speech to another does not
seem to be sudden, but graduaL In certain sections ofPi cardy,
before and during the Second World War, the linguistic
situation could be characterized as follows: when first con-
fronted with strangers, civil servants, holiday-makers from
Paris, and the like, local peasants would use a sort of local
French close enough to the average, if not to the standard
form, to be generally understood. This, they would even use
for a time among themselves in the presence of strangers,
but very soon features from their dialect, would begin to
crop up: etait, mangeait would gradually yield before letol,
Ima301 with the local 1-01 ending in the imperfect; French
vocabulary items would persist a little longer, but eventually
such a form as chaussette, pronounced locally as (in
dialect.), would be replaced by the regular Picard Ik0 Jex],
The outcome would be a form generally impervious
to anyone except the initiated. This type of stratification
suggests the possibility of a gradual word-by-word and form-
by-form elimination of dialect. through stages where dif-
ferent styles would coexist just as they do, for instance, in
unilingual French. Such a type may have been, or may still
be, fairly general along the periphery of the Basin, but
it is by no means the necessary intermediate step between
dialect. situations and unilingualism, since a sudden break is
just as likely.
Dialectal bilingualism its neater form, i.c. one with
clearly distinct dialect, and dialects, deserves to be dwelt upon
a little further, since professional dialectologists, who are
mostly antiquaries looking for old forms in any context (dia-
Icct, or dialectj), have not been inclined to describe the
socio-linguistic settings in which they operate.
The illustrations that follow are based upon observations
carried out in Franco-Provencal territory, more precisely
in Savoy, a region where dialects, were doomed as the
result of the contacts established in the course of the First
World War. Although we shall use here the present tense, the
situation referred to is, generally speaking, that which pre-
vailed at the eve of that war. Dialect! is, with careful speakers,
a fairly good approximation to average French, closer prob-
ably than in most unilingual rural regions. This largely
results from the fact that people are aware that what is good
for dialect (dialectj) is not good for French; French, of
course, is taught at school, and school attendance is universal.
Only very few, usually feeble-minded, old women speak only
dialect, although they understand spoken French (dialectj),
Dialect, is a variety of that type of Romance called
Franco-Provencal whose domain, limited byProvencal in the
south and merging gradually into the Burgundian and
Frane-Comtois areas in the north, spreads from the upper
I 17
area to
LINGUI8TIC VARIETY
coast. It is still a matter of and when this par-
ticular segment of Romance-speaking area acquired the
measure of which must be postulated by anyone
who 0 has discovered, behind the motley of contemporary
patois, a bundle of fundamental likenesses. Dialect varies
from village to village, at times from hamlet to hamlet and
o '
Its use secures unimpeded communication within a radius of
10 15 at best. Beyond that distance, peasants may
easier to use French (dialect.}. In a period ofincreasing
?f labour and geographic mobility, this means that
dialect. IS doomed: people speak French better and better
and are. never .gi:ren a chance to eliminate pre-existing or
developing variations among dialectsg; the range of forms
and words that are understood and used if necessary, con-
stantly decreases: at Hauteville, I former generations accepted
for 'bull' the form Ibar'ral beside the local Italre/; younger
people who French with those for whom Iborlral is
the normal dialect, form never get acquainted with it. Had
both parties tried to converse in dialect., both forms would
have retained their wider currency. As a matter of fact,
any reference to a bull among people from the Ibarlra/ and
the Italrel zones will, from then on, be made by means of
pronounced locally (dialect.) [tc'ro] (as
agamst Parisian [touoj), It might be advantageous to reserve
in technical parlance the word 'patois' for the designation of
minimal linguistic communities limited to a few thousands
and even a few hundreds of rural speakers. But since then
0 0 ' ,
no patois situation IS Iikely to endure, it may not be advisable
to keep a word such a transient phenomenon.
Dialect., alias local French, contains permanent and uni-
versal devia:ions from average or standard French, most, but
not necessarily all, of them carried over from dialcctj: alleren
,I Cf. Martin:t" 'Des.cription phonologique du parler franco-provencal
Hauteville .(SavoIe) , xv, pp. I-85, revised and reprinted in La Descrip-
tion phonologique avec application au parler franco-prooencal d' Hauteoille (Savoie)
(Geneva, 1956).
LINGUISTIC VARIETY 116
118 LINGUISTIC VARIETY
LINGUISTIC VARIETY
champs les naches 'grazing the , for stanuaro
baufs, quand moi 'when I did', for standard 'en mime temps que
moi, poche 'ladle', (dialect, I'p08e!) for standard louche, &c.
Most of these are normally used by both unilingual dialect!
speakers (teachers and priests, for instance) and bilinguals.
But, at any time, dialect., forms may, in the speech of the
latter, be transferred to dialect. with just the necessary
phonological and morphological adaptations. Children who
pride themselves on their early handling of the standard
language are particularly apt to effect this type of transfer:
as a calque of dialect, [a'badda 'bje- le 'plo-ts pg ka-lbo- 19
gJ1Aa], approximately 'take a big step to get over the puddle',
a ten-year-old girl said once: Abade bien les plotes pour camber
le goillat [abad bje le plot pur kabe 1goja], the phonology of
which was that of dialect, (local French), the grammatical
cement and the adverb bien were average and standard
French, but where the four lexemes abade, plote, cambe, and
goillat were exclusively local. The closest standard French
equivalent would have been Ecarte bien lesjambespour enjamber
la flaque. It should be pointed out that this utterance, when
heard, excited the merriment and the derision of the by-
standers, with some comment to the effect that if the girl
could not speak French better, she had better use the patois
which she would have handled to the general satisfaction.
What deserves particular consideration here is, however,
less the extent to which a bilingual person may let one
language influence the other, than the fact that the sentence
with its local lexicon is readily and unmistakably identified
as French, ludicrous French indeed, but French, and not the
local vernacular, because the phonology and the grammatical
monemes are those of dialect., not dialectj,
It would be interesting to determine whether some local
words such as abader, plote, camber, and goillat survive to date
in the speech of ten-year-old children who, in the regions
concerned, are by now definitely French unilinguals. If
some do, it probably will not be for long.
process we
has certainly not yet affected the whole of Romance-speaking
France, but it cannot be doubted that unilingualism is estab-
lishing itself in more and more rural districts at the expense
of Provencal, Languedoc, Gascon, Catalan, and Corsican
dialects. As regards .Catalan and Corsican, it is often felt
that their status differs from that of Gallo-Romance dialects
because, diachronically speaking, they belong together with
Hispanic and Central Italian Romance respectively. What
may count, socio-linguistically, is the fact that the use of
Catalan and Corsican dialects enables some Frenchmen to
establish linguistic contacts with foreigners with whom
French is no recourse, i.e, people who do not have French
as a common, official language. Inasmuch as these contacts
across the Spanish border or the Tyrrhenian Sea are lively
and profitable, they may retard or prevent the crumbling
down of the dialects in question into a multitude of mutually
hardly intelligible patois, the prelude to their eventual
elimination. To what extent contacts across the border do
consolidate the Catalan dialects spoken on French territory
and make them more resistant to the disintegrating pressure
of the official language than their Languedoc or Gascon
neighbours, remains to be investigated.. In the case of Corsica,
distance and the lack of territorial contiguity may largely
contribute to the conservation of the vernaculars. In those
margins of France where non-Romance forms of speech are
in common use, the possibility of using them for contacts,
across the border, with people who do not speak French
must also be reckoned with as an asset for the vernacular.
This undoubtedly plays a role in the case of Basque, with a
majority of Euzkarian speakers on the Spanish side of the
Pyrenees and a lively intercourse from north to south, parti-
cularly in the region close to the Bay ofBiscay.. Breton, spoken
exclusively on its land's end by people who all learn French
at school, does not find any support outside and must be
receding rapidly, even on the rural front" in spite of the
120 LINGUISTIC VARIETY G UISTIC VARIETY 121
I Cf. the thesis of Suzanne Sylvain, Le Creole haitien (Wetteren, 1936).
2 Best shown by Beryl Loftman in an unpublished ColumbiaUniversity
M.A. dissertation in 1953.
3 Cf. the debate in Word, with Douglas Taylor, Robert A. Hall, Jr., and
Uriel Weinreich as xii (1956), pp. 407-14; xiv (1958), pp. 367-
79; xv (1959), pp.
>.J'U....... 'V''U' ............ ~ were as thorough
differs from any 1J.L\\"'V!.il...JL'l-.l!.V..L.JL
the sense that no "I:TA1i"'''I''1lr:,,,,,, ...
larity on the
habits. There is
and
of
of features." .............i. ... ~ , .. "- .................
the a French Creole be con-
sidered a form of speech, and in
view of of the vocabulary, the
latter very a Yet if
we insist on as French the sentence abade bien les
plotes pour camber le lexemes are Franco-
Provencal, on account the use of French phonology
and grammatical argue we could call
a Creole an What is decisive, however, on
a pragmatic is the fact that the abade bien _.. utterance
was meant to and the that the speaker
had such a of French lexicon prevented
it from being as SUCll. In the case of
Creole, any utterance is meant to be Creole and nothing else,
whatever the origin of the ingredients may be. It is
never to Ewe or a lexicon.
be from attempts to force the
C111T'llI",t"l1r'\"li'"ll into previously estab-
From a standpoint, it
~ ~ ~ 1 ~ , ~ ~ ~ M France in
""r'\"I'Y'1ln,"I"li'"llIf.::'C" so much simi....
of a intel...
lectuals. Flemish should, in principle, stand a good chance
of resisting the pressure of French, since its French domain
is nothing but a fringe of a large area including north-western
Belgium and the Netherlands. Yet as long as French remains
a prestige language among the Belgian bourgeoisie of West
Flanders, the Flemish dialects of France cannot withstand the
pressure of the official language. The Alsatian situation, which
also applies to the northern stretches of Lorraine, is, in fact,
trilingual, with German dialects as vernaculars, and two com-
peting national languages, the former official language,
German, still extensively used by the Churches and in part of
the press, and French, which is by now the universal teaching
medium in schools. Both German and French could be said
to be used under the form of dialects., The local dialects,
are widely divergent, which would seem to bear witness to an
early use of a common standard, namely German, but their
position does not seem to be really threatened so far.
What seems decisive for the comportment of speakers of a
local form of speech in conflict with a standard language is
less, as generally imagined, the degree of similarity or dis-
similarity of their structure and vocabulary, than whether
the standard language is the same for all the speakers of a
given vernacular or whether the vernacular is felt as a link
with people who could not be reached through the language
one is taught at school and trained in in the army. Accord-
ingly, the socio-linguistic situation should be much the same
in Romance Gascony and Celtic Brittany, differences in
sentimental reactions notwithstanding,
The preceding rapid survey of linguistic variety in France
would not be complete unless something were said of the
language situation in the French West Indies, a few islands
which are, politically, a part of Metropolitan France. Their
population, predominantly African in origin, are taught
standard French at school, but normally speak among them-
selves what is dubbed a French Creole, namely a form of
122
LINGUISTIC VARIETY LINGUISTIC VARIETY 12
3
say
never dare put on paper.. Consequently, written style is not
spoken style. it is not always easy to distinguish, here,
between style and language: did Gregory of Tours, when he
wrote his Historia francorum, some time in the second half of
the sixth century, use a learned style of his own language, or
another language? We shall be tempted to speak of one and
the same language as long as the different linguistic forms are
felt to be complementary because each situation requires one
definite form and excludes all others, so that the user will
never be faced with the necessity of making a decision as to
what form to choose. No one will hesitate to speak of different
styles if there actually is at the user's disposal a whole gamut
of different linguistic forms merging gradually into one
another, as is practically the case with the contemporary
national languages in which anything can be found in print,
with a tendency for frontiers between genres to get blurred.
It is often said that some difference between -spoken
language and written language is rendered inevitable be-
cause so many decisive features of speech are not transferred
to writing. Whereby people mean the accentual hierarchy,
so imperfectly rendered by the occasional use of italics,
intonation, whose salient features only may be suggested by
our punctuation, and all those peculiarities that characterize
individual elocution. This amounts to saying that writing,
with print as its ideal form, is a set of discrete visible symbols,
each corresponding to some discrete audible unit of speech,
and anything that is not discrete will be sacrificed in the
transfer.. There is a large measure of truth in this: in the last
analysis, many of the features that differentiate written style
from spoken style can be traced back to a need, in writing, to
compensate for the loss of suprasegmental and individual
elements of speech. Through too much insistence on the
blurring of contextual features, however, one is apt to forget,'
or at least to minimize, the importance of situation in spoken
communication, and the necessity to compensate for its
"' ... .!I..!I.."-"'-il '''-<''V and as it is
of Basses-Pyrenees, Creole users would achieve a complete
mastery of French far more easily than their Euzkarian
fellow citizens.
The widespread beliefin the unity and homogeneity of the
language of a given 'community' does not only conceal the
linguistic variations existing within the boundaries of each
state, but tends to convince even the educated that the
language people speak and the language they write is neces-
sarily the same. This is, in some cases, palpably untrue, and,
in others, open to question. When a linguistic community,
illiterate hitherto, gets acquainted with the art of writing,
this is used for the rendering of some other language. In such
a case, dissociating the writing from the language requires
a power of abstraction that may be absent, so that the first
persons who write do so in the foreign language. This may
become institutionalized, so that literate people go on speak-
ing their vernacular, sometimes exclusively, and do not
know how to write except in another language. In many
cases, the written language is that of a 'classical' literature or
that of a liturgy, which was the case with Latin in medieval
Europe, and is still so with Sanskrit in India and Koran
Arabic in Muslim countries. This, it is true, is not always
exclusive of efforts to apply writing to the vernaculars. But,
as a rule, those efforts take centuries to bear appreciable
results..
It could be objected that the fact that some people use a
foreign or ancient language as a written medium is a very
special case and that it does not mean that the written
language of a nation is necessarily different from its spoken
medium. Most people would concede, however, that there
may be, between one and the other, a difference of style: the
chief reason why people do not write as they speak is prob-
ably that since writing leaves permanent traces, whereas
speech, unless recorded, is lost for ever, writers are far more
12
4
LINGUISTIC VARIETY
LINGUISTIC VARIETY 12
5
absence in : a speaker may at all times refer to what he
has, hie et nunc, common with the hearer; the author has
nothing in common the reader but his text, and the fact
that they belong to the same linguistic community. It will
not do to argue that the author constantly re-creates situa-
tions so that he will use 'shifters';' words that are understood
in situation only, like I, you, here, lastyear, just as easily when
he writes as when he talks, very fact that he has to
recreate his situations if he wants to make use of a large part
of the lexicon suggests that an appreciable part of his activity
is devoted to descriptions and presentations from which
'shifters' are excluded except when, quite exceptionally, the
author ('I') addresses his readers ('you') .. As a matter of fact,
contemporary literature has done much to introduce spoken
style in written matter: the ideal of some writers is obviously
to try to do away with presentations and descriptions and to
let their characters 'speak for themselves' .. But the difference
between written style and spoken style is not thereby wiped
out: they speak in print, instead of speaking aloud, and the
ideal of a written style, self-sufficient, relying exclusively on
discrete, linguistically central, non-expressive elements, sur-
vives, less perhaps in literature in the narrow sense of fiction,
than philosophical and scientific writings.
The distinction between literary language, or style, and
colloquial speech is, of course, not identical with that be-
tween primary spoken form and secondary written form, as
our use of 'written' and 'spoken' style may have suggested:
contractions like don't, can't, ain't belong to various levels of
'spoken' style; still, they have a written form, and the many
words are used outside of written texts have a
pronunciation of their own, if only because they may have
to be read aloud. Still, there are a few words that are so
definitely spoken rather than written that there is no tradi-
tion governing their spelling; such is French jpagajj, 'mess',
I So called by Otto Jespersen, Language, Its Nature, Development, and Origin
(London, 1922), pp. 123-4.
which, in may appear as pagaye, pagaie, or pagaille;
popular German I'fuzelj, 'spirit', cannot even be given a suit-
Fusel, is in rJ.I "."...,.r'\.",.."I.:>.l'l
would correspond to *I'ffizdlj and *Fussel to *I'fusdlj. The
reverse is far more common: hosts of words are so
generally learnt through print no single pronuncia-
tion can permanently establish itself (gerrymander with [g] or
with [d3]?).
In a comparison of the frequently distinct systems pre-
sented by the written form and the spoken form of one and
the same language, one may, in order to make the contrast the
more striking, combine in one pattern the archaisms of
the spelling and those of the literary style on the one hand, the
innovations of the spoken form and those of colloquial speech
on the other hand. As an illustration of how profoundly
different things can be subsumed under a single .language
label, the case of French may be dealt with at some length.
English spelling gives, in many respects, a less reliable
picture of the spoken form than French does: it is bad enough
for native speakers who have to discover that what they pro-
nounce jredj is either redor read; it is a frightful nuisance for
foreigners who have to pronounce read sometimes as jridj,
sometimes as jredj; but it does not give an inaccurate picture
of the main lineaments of spoken grammar. French spelling
gives foreigners fairly reliable hints as to how to proIlounce
vowels; where it is under-differentiated (jaj or [a], jej or lei,
and the like), Frenchmen as a whole do not really care about
differentiation in speech. The real trouble is for French-
men who waste the best years of their childhood trying to figure
out they shall be content with the bare stemor when they
shall write -s, or -nt, or -t.. Formerly, those children who went
to school had to learn grammar as a preparation for Latin
lessons to come.. In the world of today, where hundreds of
millions of children go to school with no intention of ever
learning Latin, grammar, a highly abstract affair, perfectly
unsuited for the budding intellect of the average ten-year-old
LINGUISTIC VARIETY
youngster, is doomed as a universal prerequisite. French
children, however, will have to go on carrying the old burden
as long as they have to decide when liJatl is to be spelled il
chante and when ils chantent.
What follows is a sketch of the verbal forms of French in
what could be designated as 'unguarded' spoken usage. I
This, in turn, may vary from one region to another, and what
will be considered specifically is the type ofspeech observable
among people living in Paris. But it must be understood that
this type has a much wider currency and that speakers of all
provinces tend to conform to it. Some of the forms below
may, in their written garb, strike one as extremely vulgar and
belonging to a popular rather than a familiar level. Yet all of
them can be heard in the speech of all social classes. They
are normally used by the present writer in his home.
Some forms that have been disregarded cannot be said
to be totally absent in informal conversations: the present
writer, even in his most slipshod linguistic behaviour, may
well be pushed into risking an imperfect subjunctive in a cas-
cade ofsubordinate clauses. But he will hate himselffor being
tricked into it, since he isjust as likely to use a wrong I-sl form
in the third person singular as the right, I-s/-Iess form. As
a matter of fact, our examination will be restricted to those
forms which adult speakers of any social class know how to
use and use without tremor.
By 'verbal forms', we mean the set or sets of verbal signi-
jiants, with their variants, that children have to learn before
they can handle the language in a way that will satisfy their
elders. Therefore it is not our intention to dwell here on
compound forms as distinct from their components, the
auxiliary on the one hand, the participle and the infinitive
on the other. French, just like, and even more than, English,
has a host of compound forms, even if we disregard quasi-
I For a fuller treatment of the same, see 'De l'economie des formes du verbe
en francais parle' in Studia philologica et litteraria in honorem L. Spitzer (Berne,
1958) , PP' 309-26.
LINGUISTIC VARIETY 127
modal combinations such as il doit aller, il peutfaire, and the
most striking development in this respect is the
spread of doubly compound forms such as (quand) il a eteparti
or il avait eupris, whose ultimate source must have been the
necessity of distinguishing between the true perfect with a
present meaning, as in quandj'aijini, and the compound past,
quandj'ai eujini, built from the former on the analogy ofj'ai
faim, a present with, as a corresponding past,j'ai eufaim. Any
study of the syntactic functioning of the verbal system will
have to place such a form asj'ai eujini among the constitutive
elements of that system. But, for us here, it is nothing but a
present form ai and two past participles eu andjini.
Before we pass on to the analysis of our forms, the problem
of linkings (liaisons) had better be dealt with apart. French
verbal endings are felt to be particularly tricky because,
according to whether the next word begins with a vowel or a
consonant, a characteristic liaison consonant may be sounded
or not. This is quite true for any sort ,of 'guarded' French
which is to some extent affected by the spelling. But in the
informal style we are considering here post-verbal linkings
have been preserved only in some very specific situations, so
that they should no longer be considered formal variations
of the verbal forms, but parts of variant forms of certain pro-
nouns. Most post-verbal linkings that are suggested by the
spelling are either fanciful or optional and actually excep-
tional in familiar delivery: pronouncing a I-z-j in tu donnes a
..... would sound ridiculous although the liaison is acceptable
after sg. 2 in the case of the copula and a few very common
verbs: contrast, for tu es une ... 'guarded' ItyEzyn/, familiar
ItyEyn 4> .. .. I, and slipshod ItEyn/; ils donnent une 4> is i ~ n y n
.. 4> .. I, just like il donne une . 4> ., or l i l ~ n t y n I in careful
speech; the desire to distinguish between sg.. 3 and pl. 3 un-
doubtedly contributes to making the I-tl linking here more
frequent than the I-zl linking of sg. 2: in the same style we
may hear j-tj linking in ils donnent un livre, but no linking in
les parents donnent un livre, where the plurality of the subject
12
9
o don-ra
vu don-re
i don-ro
o don
vu don-e
i don
odon-e
vu don-ie
i don-e
LINGUISTIC VARIETY
3 don
ty don
i don
3 don-e
ty don-e
idon-e
3 don-re
ty don-ra
i don -ra
2. The imperfect, which we need not call 'imperfect in-
dicative' since we do not reckon with any other imperfect. It
is characterized by an 1-8/ ending in sg. I, 2, 3, pl. 3, and
informal pl. I .The endings are /-i5/ for less informal pl. I and
/-ie/ for pl. 2. Hence the following paradigm:
It is characterized by a zero ending in sg.. I, 2, 3 andpl, 3,
/-0/ in pl. I and /-eJ in pl. 2 .. Informal French, however,
normally employs I forms a zero /0/ as
the subject pronoun: /0 d'Jn/ on donne, instead of /nu d'Jno/,
nous donnons, with /nu 0 d'Jn/, nous, on donne, as an insistent
form. Here is the resulting paradigm:
3. The future which is marked by a combination of /-r-/
plus a vowel which is /-e/ or I-E/
I
for sg. I, I-a/ for sg. 2,3, and
informal pl. I, /-01 for traditional pl. I and pl. 3, /-el for
pl. 2. The ensuing paradigm
is etymologically connected and, perhaps, synchronically
supported by that of the present of 'to have': /e
2
a a a av-e 0/.
I Cf. Prononciation, pp. I 22-6: among Parisian speakers, 53 per cent. rhyme
seraiwith etd; most of the rest rhyme it with etait; the percentage for I-el sinks
to 40 for the youngest third. Note that I-el is widely known to be the 'right'
form.
2 It would be worth while investigating to what extent the speakers who use
leI as the sg. I of the present of avoirare the same who pronounce future sg. I
as I-reI.
811928 K
the only
are the following:
v ...... ..,,""C"1l>'YllT indicative'
"Il'"'I.V'C.'C1.c::>"D"'l'f- tense positively
"Y'I ".....11'10- 'Y'1"AClA1Yll1" tenses.
I-r-I, is
in imperfects.
In of II-! .. "',cl>Ybron
moods and tenses we
I. The present,
since it is, in 1'Y1l-tA1"1r1l"l ':)
marked as such f-h"'V'''''''"1l1oon
LINGUISTIC V ARIE
factors intervene which
queer,
r-rvn r-ro tH/011T a e 1,
.................... .,.. ...... :-. here is never obliga-
tory as item is not a If it is a
pronoun in it is a third person pro-
noun (il, ils, elle, elles, on), and these are preceded by
a I-t-I, its is a regu-
lar part of the as in il ils donneni,
donnent-ils, or an ad hoc as in donne-t-on. Only im-
peratives may comple-
among en with a
vowel; as a rnatter of fact by I-z-I
irrespective of whether the form is permanently
spelled with -s (cours, cours:)) or not (va, vas:)). If we leave
the spelling out of the we may all this
as follows: the li(l), are pro-
nounced Iti(1), ti (1), tel, t5I after the
verb this is not common in current speech
where interrogation is indicated by means of lesk/,
est-ce que, rather by ; the (ai, en, and
li/,y, are pronounced Izal and Izil immediately
after an imperative. thus dealt
with in the frame of the of pronouns, they
can be disregarded in an French.
we call of verbal
form that follows as a radicaL An
ending may be of distinct
monemes: the -rais made up of
is found
LINGUISTIC VARIETY
LINGUISTIC VARIETY
It might be tempting to assume that its comparative com-
plexity has been instrumental in the promotion of compound
futures of the il vadonner Ii va danel type. But it should be kept
in mind that the present of the auxiliary aller is just as com-
plex: IVE va va va al-e vb],
4. The conditional, whose status as a tense or as a mood we
need not discuss here. It is marked by I-r-I followed by the
endings of the imperfect.. Hence:
When it is used as the future in the past (il disait qu'il donne-
rait 'he said he would give') it is often replaced by a compound
form parallel to il va donner (il disait qu'il allait donner), but the
I-r-f form is indispensable as the conditional proper: il don-
nerait cher pour. .. .. ..
5. The subjunctive, which corresponds to the present sub-
junctive of traditional grammars. Its marks are zero for sg. I,
2, 3, informal pl. I and pl. 3, I-iol and I-iel for less informal
pl. I and for pl. 2 respectively. The informal paradigm
is identical with that of the present with the exception of
pl. 2. Only such verbs as make use of different stems for pre-
sent and subjunctive manage to keep the two tenses distinct
throughout. The subjunctive is actually in most cases what
has been called a 'grammatical servitude': its use is largely
restricted to certain grammatical contexts from which other
tenses are excluded: after il faut que, vous donne; and vous
donnere; are excluded; the subjunctive form vous donnie; is
compulsory and consequently does not contribute any in-
formation. But children learn frequent forms like je sois, je
fasse very early, so that these survive after young speakers
have learnt to use analogy, and the subjunctive shows few
signs of obsolescencee
6. The imperative has only three distinctive forms, those of
sg. 2, pl. I and 2, with zero, 1-01 and /-e/ respectively, i.e, the
same endings as the present. Traditional pl. I imperatives in
/-0/ seem seriously affected by the disaffection for such forms:
traditional chantons, 'let's sing', is likely to be replaced in
many situations by such phrases as allez! on chante or allons!
on chante, in which the status of allons is close to that of an
interj ection.
The non-personal moods present three distinct forms ..
those of the infinitive, the present participle, and the past
participle. The present participle is always in I-a/, but the
infinitive and the past participle have endings that vary
from one type to another: most of the verbs that present the
same stem throughout have /-el for both (/dan-e/); those that
add their personal endings to different stems end their in-
finitives in I-rf (/fini-r/, Ibat-r/), hut in the case of such verbs
as deooirkivixes] or partir/partir/, we may either decide that
the ending is just I-rI and the infinitive stems are Idvua/ and
/parti/ distinct from both /dua/, /par/ ofJe dois Je pars, and
/dv-/, /part-/ of vous deoez; vous parter, or posit that the infini-
tive endings are /-uar/ and /-irl respectively, the stems being
in that case those of present pl. 2 /dv-/ and /part-/. Past parti-
ciples of verbs with /-r/ infinitives may either be said to end
in zero (/fini/; cf /3 fini/}, in I-i/ (/part-i/; cf /3 par/) or in
/-yI (/bat-y I; cf /3 ha/), or be all of them considered bare
stems (/parti/, Ibaty/, just like lfini/).
But the infinitive and the participles are, of course, quasi-
nominal forms that stand aside from properly conjugated
ones. The past participle and the infinitive are in French so
frequent that children often learn them as if they were in-
dependent words with no formal allegiance to the personal
moods. This accounts for the fact that obsolescent French
verbs are as a rule verbs which people do not know how to
o dan
vu don-ie
i dan
o don-re
vu don-rie [donorjc]
i don-r e
3 don-rs
ty don-r e
i don-r e
bered and used if necessary.
14 in consequence, we dissociate conjugated from non...
conjugated forms and concentrate on the former, we may
summarize our observations by saying that in spoken collo-
quial French all verbs are inflected in the same way, i.e .. by
means of the addition of the same endings to a stem, but that
whereas most verbs, such as Je donne (/3 d'Jn/), present the
same stem throughout, others distinguish between a short
stem (e.g. 13 ba/) and a long stem (/vu bat-e/), others still,
among those with the highest frequency, presenting short
stems and different long stems with peculiar distributions
(e.g, 13 VEl, Ity val, 13 i-t e], Ivuz al-e/, 13 aj/, for aller). Such
isolated remnants as Inu s'Jml nous sommes or Ivu fet/ vous
faites, where it made sense to identify an l-iu] ending and
a I-tl ending as long as the simple past, with its regular I-ml
and I-tl endings in pI. I and 2 (/nuz ala... ui] nous alldmes, Ivuz
ala-t/ vous alldtes) was part of the system, must, in the frame
of contemporary colloquial French, be considered unanalys-
able amalgams.
The picture is completely different if we consider now
literary French in its written form. There is probably little
need here to list forms that can all be found in any grammar
of French. Two tenses, the simple past and the imperfect
subjunctive, that have been disregarded in the preceding
survey, re-enter the picture, and thereby the connexion
between the quasi-nominal items and the conjugated form
is largely restored: donne is predictable from Je donnai, parti
from Je partis, couru from Je courus, and vice versa in spite of
exceptions like Je cousis, cousu. But the main difference lies
in the fact that all verbs do not take the same endings: some
take -e, -es, -e in the singular of the present indicative, while
others take -s, ... s, -t. It is therefore necessary to posit different
conjugations, just as in Latin, Italian, or Spanish, and this is
actually what we find in traditional grammars; irrespective
of whether grammarians decide to list two of them, if they
start from the or more than two if the
infinitive, the form under the verb is quoted, is taken
as a one is said to
be productive, the one with endings in -e- (-e, -es, -e, -er, &c.) ..
It corresponds roughly to the one-stem verbs of spoken collo-
quial French, coincidence between productivity and
invariability of the stem is, of course, not fortuitous: a
denominative canoter from canot or a loan like dribbler
from English dribble cannot be expected to modify its stem,
unless the modification is somehow automatic and conse-
quently predictable; e"g./3 driblj with the stem pronounced
as [daibl], future 13 pronounced [dibaloae]. But
this does not lessen the contrast between the two systems, one
with distinct conjugations, the other with a single conjuga-
tion, but a distinction between those verbs whose forms can
all be predicted if anyone of them is known, and those for
which this is not the case ..
Two so fundamentally different systems can be made to
coexist as two styles of the same language only through a
painful and protracted training, which a nation will impose
on all its new members when it is felt that thereby a precious
cultural heritage will be preserved.' This is not the place to dis-
cuss whether the burden could be lessened without affecting
that heritage, for instance through some drastic spelling
reform. OUf only aim in presenting the two systems was to
illustrate how different two linguistic forms that purport
to be 'the same language' can be.
LINGUISTIC VARIETY
comucate. .......... +- ""Thl''''''''''''' ......." ..'"Y'11t- ... TT.c>. and are remem-
LINGUISTIC VARIETY 1
33
LINGUISTIC EVOLUTION
135
LINGUIS r c EV LU ION
I
T is customary, among contemporary descriptivists, to
consider that pre-structural linguistics dealt exclusively
with the history of languages. This is not quite accurate:
traditional linguistics, as practised during the last 150 years,
has probably more often been engaged in comparing genetic-
ally related languages and trying to account for some feature
of one by reference to some features of another, than in trying
to determine how and why a given language had evolved
through the centuries. In other words, scholars have been
more inclined to point out correspondences than to explain
them. It is true that no serious research can be pursued that
is not based upon observation; history implies the study of
documents, and, even for the best known among languages,
our documentation is full of gaps. Therefore, it may have
been scientifically safer in many instances to avoid historical
treatments altogether.
Actually, most linguists, until the dawn of structural lin-
guistics, were not aware of any necessity of distinguishing
between diachrony and synchrony, and, quite frequently, they
studied what Saussure called etats de langue without drawing
any clear boundary between comparative attempts and dia-
chronic references on the one hand, and synchronic obser-
vation on the other hand.
Today, after decades of conscious synchronic practice, it
is certainly easier to understand what historical linguistics
really implies. On the plane of general linguistics, it amounts
to determining how and why languages change through
time, and this is what is meant here by linguistic evolution.
This problem was at one time a favourite with some
language theorists, but mainly as far as sound changes were
in question. Sober scholars, in spite of their reluctance to
enter the realm of hypotheses, had had to make up their
minds regarding the nature of phonetic changes, and, in
order to reach a decision, they had been compelled to con-
sider the problem of how the obvious, if not total, regularity
of sound changes could be explained. This regularity, whose
reason is obvious as soon as speech is found to be analysable
in terms of a definite number of discrete units, the phonemes,
was for several decades the subject of strenuous debates,
which resulted in its being generally acknowledged, at least
as a working hypothesis. But since speech sounds were at that
time usually considered not to belong to language proper,
people were tempted to look for causes of change that lay
outside of language, and preferably outside of man; hence
the ever-recurrent references to climate, latitude, altitude,
and the like.
We have now identified the outward manifestation of
language as a perfectly legitimate part ofit, and consequently
we are not inclined to account for its vagaries in any other
way than the one we shall use for other linguistic changes. It
would seem that if languages change, as we know they do, it
is, basically, because the needs of their users change, and
this has been found to apply to phonology as well as to lexicon,
morphology, or syntax. This, of course, involves a total
revision of traditional views regarding sound changes.
Another basic contribution of contemporary research to
the problem of linguistic change is the establishment of a
relation between the frequency of a linguistic unit and its
form, so that a change in frequency may involve a change in its
phonic aspect. This, a consequence of the theory of selective
information, applies to units of both articulations, to phon-
emes as well as to monemes. It leads to the conclusion that
since the rise in the frequency of a unit is nothing but a rise
in the number oftimes that unit is used, any use ofa linguistic
unit contributes, even if only in an infinitesimal way, to an
increase in its frequency. This automatically entails a lower-
ing of its contribution to information and a tendency to
LINGUISTIC EVOLUTION LINGUISTIC E VOL UTION 137
reduce its cost, that is the output of energy it requires for its
memorizing and production. Such a reduction is likely to
physical aspect of the unit. This
could be summarized by saying that, strange as it may sound,
a language changes because it is used.
People as a rule are not conscious of any change taking
place in their language. When they are told and shown how
different it was four or five centuries ago, they are apt to
wonder how this came to pass and imagine some period of
rapid change from one period of stability to another, the
one they imagine they are enjoying. It is not difficult to
understand the reasons of the common illusion that one's
language is stable and homogeneous: people tend to identify
language and its written form, and would naturally think that
nothing changes as long as spelling has not budged; as a rule
they do not and cannot remember how they spoke ten or
twenty years before; everyone is used to and considers normal
many forms and turns which he himselfnever uses, but which
he seems always to have heard; he is likely to brand as
'accent' any deviation to which he is not used or which is so
considered in his town or province; but as regards other
features, his tolerance is boundless. This is all to the good;
the chief aim of language is communication; it is greatly
advantageous that on the one hand something that hampers
communication should be resisted and denounced, and on
the other hand that something that does not should be dis-
regarded or condoned.
Yet no one will deny that under the pressure of technical
advance the vocabulary of a language is likely to expand;
most of us remember a period .when radar and sputnik did
not exist. We may also at times become conscious of the
obsolescence of a word which was once of daily occurrence.
But it is more difficult to imagine how the changing needs of
man or his communities may influence the more intimate
fabric of his language, namely the morphology and, above
all, the phonology.
order to understand how the influence of technical,
economic, or cultural changes may spread to the inner core of
language, we have to remember what we do when we want
a linguistic communication to be more specific. If I want
someone to pass me a certain book and if the request 'give me
the book' does not make me get the desired object, I shall
try to be more specific and say give methe blue book or givemethe
blue book that lies on the far endof the shelf I might also replace
book by another in itself more specific term like octavo. But it is
far more usual to achieve specification through the addition
of new elements to the sequence of meaningful units than
by replacing one term by another. Besides, in the last analysis,
a highly specific term like octavo is nothing but the residue of
a phrase formerly used as a specification of book or some
equivalent. People do not, as a rule, invent new monemes,
All this is in keeping with the analytic process I have called
the first articulation, according to which experience is sub-
jected to an analysis manifested in linear form.
Now ifmy request had been presented in a very simple and
somewhat primitive household, there would have been no
need for any specification, because there would have been
only one book there, namely the Bible or some almanac.
But life having become more complex, people need more
books and more specific books. On the plane of language, the
result is that, while I might have got along in the past with a
simple utterance made up of the injunctive give accompanied
by two handy complements me and the book, the satisfaction
of my needs will require now the addition of a relative clause
and that of an attribute blue, which is an original predicate
used here with the same function as that of a relative clause.
It is thus clear that an increasing complexity of social rela-
tions will be accompanied by an increasing complexity
of syntax. Division of labour will involve the appearance
of new forms of human and material relations which
will determine the appearance, in language, of new func-
tions.
I The basic treatment of least effort, also in reference to language, is still
George K. Zipf's Human Behavior and the Principle of Least Effort (Cambridge,
Mass., 1949).
language and through a limited period, the various conflicts
and trends in the frame of the permanent needs of human
communication.
In order to understand how and why a language changes,
the linguist has to keep in mind two ever-present and antino-
mic factors: first the requirements of communication, the
need for the speaker to convey his message, and, second, the
principle of least effort, which makes him restrict his output
of energy, both mental and physical, to the minimum com-
patible with achieving his ends.' Now the objection is often
raised that human activity in general, and linguistic activity
in particular, may in many instances be an end in itself:
a play: talking, more often than not, contributes very little
to information; many people talk because they like it, not
because they have anything to communicate. But this does
not imply that linguistic evolution is not determined by the
principle of least effort. Talking is often just a game, but a
game is only worth while if he who plays the game sticks to
the rules. Cheating makes sense only if the game ceases
to be an end in itself, and there is, for chatterboxes, no reason
why they should cheat in the game of talking. The rules of
the game of speech are laid down by its communicative uses.
At every stage, the structure of language is nothing but the
unstable balance between the needs of communication,
which require more numerous and more specific units, each
of them of comparatively rare occurrence, and man's inertia,
which favours less numerous, less specific, and more fre-
quently occurring units. It is the interplay of these two main
factors that constitutes the essentials of linguistic economy.
We shall therefore concentrate on language as a communica-
tive tool, since this use of language gives it a form likely to be
imitated in all its other uses. We may thus posit, as the basic
principle of language economy, that the amount of energy
138 LING U 1STICEVOLu T ION
A language like Latin, considered in itself and in its evolu-
tion towards its modern Romance representatives, affords
a fine illustration of how two types of function marking,
corresponding to two successive periods in the evolution of the
language, combined for a while, before the older type was
eventually eliminated by the expansion of the more recent
one: on the one hand, an inherited, formally rigid and strictly
limited system of case-endings no longer capable of taking
care of the expression of all the relations needed in Roman
society; on the other hand, a set of former adverbs gradually
promoted to the role of function marking prepositions con-
stituting a very handy and expendable instrument, which,
for a while, eked out the case system, but finally eliminated
it as an unnecessary burden.
What this Latin example illustrates is not only the way
increasing social complexity determines an expansion of the
functional complexity ofthe corresponding linguistic medium,
but also the protracted resistance of the language against the
reorganization required by new social and communicative
needs. Inertia combined with redundancy delays the spread
to all the parts of the language structure of the repercussions
of some initial change. As a result of this, a language is
necessarily the battlefield of conflicting trends, the linguistic
consequences of past social changes running against the im-
plications of new innovations corresponding to new stages in
the evolution of society and coming with them to terms
which represent the structure of the language at every step.
This implies that it is extremely difficult to trace linguistic
causality back to its ultimate social antecedents. Linguists,
once they have ascertained the decisive influence exerted by
social factors on linguistic structure, should not try to do
what they are not trained to do and what might lead them
into the realm of unverifiable hypotheses, namely to examine
the details of that influence and venture into the field of
cultural history.
Their real task is to observe and describe, within a given
LINGUISTIC EVOLUTION 139
14 LINGUISTIC EVOLUTION LINGUISTIC EVOLUTION
spent toward linguistic ends will tend to be proportionate to
the amount of information to be conveyed.
might be as if no feature or unit in a
language would endure unless it contributed to communica-
tion a share corresponding to the efforts required by its
memorization and production. But this is palpably untrue;
every language carries along a heavy load ofmost uneconomic
forms which, for various reasons, have been retained through-
out centuries. Some are learnt by children at an early age,
before they are able to use analogy, when the language they
use is so poor that its economy is bound to be very different
from that of more-advanced users; others are forced upon us
by reasons of prestige; others still are preserved because no
indication is to be found anywhere as to how they could be
adapted to current needs. But the main reason why the
energy spent is not strictly proportionate to the amount of
information intended by the speaker derives from the fact
that communication hardly takes place in ideal situations;
some amount of noise, man-made or not, interferes with the
identification ofthe successive linguistic units by the audience;
besides, people often listen intermittently because they rarely
ascribe more importance to what is said to them than to their
own latent or conscious preoccupations. Depending on
various conditions, messages will be more or less repetitious,
and this will be determinant for the average handling of
language. Redundancy, in various forms, is a basic necessity
of linguistic communication. As a matter of fact, human
nature is such that a total elimination of redundancy might,
in many circumstances, entail a serious increase of the energy
spent on speech; as popular practice indicates, it is probably
easier to use double or triple negatives (1haven't seen nobody ...)
than to order one's utterances in such a way that there
should never be any repetition. Once the word yesterday
figures in an utterance it is certainly less trouble to let the
aura ofpast action pervade all that follows and determine the
choice of past tense than to take advantage of the precise
indication of the past contained inyesterday and dispense with
tense endings. The more so, ofcourse, ifwe have learned to do
so as ..., ..., .
Redundancy is indispensable for the transmission of
language to new generations of speakers: a good many words
are learned through identifying certain aspects of certain
situations, say a given animal, the horse, with a certain vocal
product. But thousands of others, which do not correspond
to concrete objects or beings, are normally learned through
redundant contexts, hungry, for instance, iflearned by a child
when listening to such utterances as he's hungry; give him some-
thingtoeat, or I'm hungry; when do wehave dinner? Learning new
words through redundant contexts is a very common occur-
rence throughout life, whether this takes place in one's own or
in some foreign language.
The importance of redundancy does not, of course, in-
validate the concept of language economy, but reminds us of
its complexity; what is easiest is not necessarily logically
simplest. It would be totally erroneous to identify informa-
tion meant and information actually conveyed.
All this means that linguists, in their efforts to understand
how language changes, should undoubtedly take advantage
of information theory, but that they should thereby keep in
mind that, in any speculation involving cost, i.e, in our case,
energy spent toward storage and production, or, if we prefer,
memorization and actual use in speech, they will have to
avoid attempts at mathematical formulation. There are two
reasons for this: first and foremost, we have no way ofquanti...
fying that energy except in a very approximate fashion; we
may point out, for instance, that, everything else being equal,
the addition of a new phoneme to a succession of such entails
the expenditure of extra energy: feed 'costs' more than fee.
Secondly, even if we could achieve some quantification,
other factors than the ones involved in information theory,
factors which might, for instance, tie up with sociology,
would remain so potent that we would not know what to do
LINGUISTIC EVOLUTION LINGUISTIC EVOLUTION 143
with our hard-won mathematical accuracy.. Therefore, in
language dynamics, we shall have to be satisfied with deter-
mining in what direction the variation of certain factors is
likely to cause that of others.. These variables are (I) the
number of the units among which the speaker chooses at a
given point in the utterance, (2) the probability of units,
which we shall in practice identify with their frequency,
either in a definite position or generally, (3) the cost of each
unit, and (4) its information..
We call information whatever reduces uncertainty through
the elimination of certain possibilities.. This means that in-
formation is not the same as meaning. If I say hehasp.. .. . . and
stop short, phas, ofcourse, no meaning, but it carries informa-
tion because it excludes the possibility that the utterance may
have been meant as hehasgiven or he has seen.. If I say he haspr
....., r has no meaning, but, again, it contributes information
because it eliminates hehaspushedor hehasplaced. This implies
that what is going to be said about the dynamics of language
applies to all linguistic units, distinctive or significant, pho-
nemes and monemes alike.
As is well known, the more numerous the units among
which the speaker may choose, the more informative each of
them will be: if someone is trying to locate a place on a map,
and it: in my directions to him, I am only allowed to use the
four cardinal points, telling him to look south will limit his
research to one-fourth of the expanse; but, if I may also use
compounds such as south-west and north-east, his research
will be limited to one-eighth of the map. His uncertainty will
thus be reduced, and we may say that each one of the direc-
tions in the eight-point system is more informative than each
one in the cardinal, four-point, system.
Now if, in a certain place, the wind blows from the west
just as often as from any other direction, the information
contributed by west in reference to the wind will be that of
each unit in a two-unit system; in other words, it will be
smaller than if north, south, east, and west had been equally
likely .. This means that the information contributed by a
unit depends on its probability in the context where it
occurs. more probable a less informative it is..
The probability ofa unit is presented in terms offrequency.
The relations between number of units or frequency of
units on the one hand and information on the other hand
follow, as it were, from the very definition of what informa-
tion is. They are necessary and,should we say, automatic.
If the frequency of an item decreases, its probability must
necessarily decrease too, and nothing can prevent its informa-
tion from soaring. The relationship between information and
cost is of a totally different nature; if the information of a
unit is affected either because of some increase or decrease
in the number of units with which it is in competition, or
because of some change in its frequency, we may expect that
speakers will be tempted to let the output of energy in their
production of that unit be partially determined by its new
informational role. In other words, people will be ready to
pay more for more information to be conveyed and inclined
to pay less if information decreases. But it is by no means
certain that they will, in all cases, be in a position to yield to
that temptation: slurring, incase of a decrease in informa-
tion, seems excluded as long as the unit in question, a word,
for instance, is still in competition with a large number of
similar units which have to be kept distinct. Only where the
choice is limited to two or three possibilities, as when only
yes or no can be expected, will the barriers raised by phone-
matic articulation crumble down, and grunts such as uh-huh
take the place of a neat succession of distinctive units. If
phonemes cannot be tampered with, the number of them
may, in specific instances, be reduced through abbreviation,
as when professor becomes prof and underground railway, under-
ground, or through replacement, as when tube replaces under-
ground. But there are cases and situations where, for various
reasons, nothing can be done: it seems that, in America, no
one has been able to find a handy and generally acceptable
144 LINGUISTIC EVOLUTION
LINGUISTIC E VOL UTION 145
substitute for the comparatively frequent elevator-operator, and
there are many situations where a professor is always a pro-
fessor and never a prof. If, now, the information of a word in-
creases, i.e, if its frequency diminishes, the bulk of its signijiant
may well be preserved unchanged: the French for 'spinning-
wheel' is still the three phoneme rouet,just as it was in the days
when the word was in daily use, and this is likely to endure as
long as no homonymic conflict arises.
Statistically, it seems true that, in all languages, the sum
of the phonematic segments needed for the 500 most frequent
words is smaller than that obtained for the next 500 lexical
items; which means that frequent words are, on the average,
shorter than rare words. But if the relationship between fre-
quency and cost were of the same type as that between fre-
quency and information, we should be able to state that, for
a given frequency, a: word should have n phonemes, no more
and no less, and that a word of n phonemes should definitely
belong to a given frequency range. This is, of course, not the
case; conspicuous with its four syllables and eleven phonemes
is so much more frequent than dinosaur with three syllables
and seven phonemes. We have here a clear indication that a
strictly mathematical treatment of the problems of language
dynamics is not practicable. But the unmistakable existence
of an inverse relationship between frequency and linguistic
complexity is a most precious discovery.
This applies to all linguistic elements in so far as they con-
vey information and to distinctive non-significant elements as
well as to meaningful units, and it is' perhaps advisable to
illustrate this relationship, first by a reference to some phono-
logical phenomena: the advantage of phonological illustra-
tions lies in the fact that everything is so much simpler there,
with a definite number of phonemes per language and no
necessity to reckon with meaning. This would seem to out-
weigh the disadvantage resulting from the fact that many
philologists feel more at ease among meaningful than among
distinctive segments, among words than among phonemes.
It is quite that the of a language
are of different degrees of complexity, and that this should be
one to understand
dynamics of a phonological system. I But when dealing with
phonological complexity it is easier andsafer to lookfor it along
the line of successivities and operate with the assumption
that two successive phonemes require more energy than one. It
is clear, on the other that two successive phonemes
give far more information than one: we get much closer
to I have praised if we perceive, after I have, two phonemes
JprJ instead of just JpJ. come to the same conclusion if
we start from JprJ as a whole whose frequency we compare
with that of, say, JpJ. We may assume that, in a corpus of
English utterances, there would be about twenty times as
many Jpl's without a following JrJ than Jprl's, which means
that the information conveyed by JprJ would be considerably
higher. Therefore, it is worth while spending more energy on
JprJ than on JpJ. Now, let us assume that, for some reason or
other, the frequency of JprJ increased so as to become equal
or comparable with that of JpJ. Speakers would certainly be
tempted to reduce the amount of energy devoted to the pro-
duction of JprJ and make it similar to the one used for JpJ.
But dropping JrJ would not do, because JprJ and JpJ would
then be confused and, on account of the high frequency of
both, misunderstandings would be so frequent, that speakers
who, in principle, want to be understood, would correct them-
selves at once, so that JprJ would never get a chance to be
reduced. A reduction could only take place if either JpJ or
JprJ could be induced to become something else without
creating conflicts: if IpJ for instance, become JfJ with-
out conflicting with any previous labial fricative, then JprJ
could become JpJ without creating any confusion.
Now a sudden spate of Jprl's is something I have never
heard mentioned in any language at any stage ofits evolution.
I Cf. G. K. Zipf, The Psychobiology of Language (Cambridge, Mass., 1935)
and Economic, pp. 132-8.
811928 L
LINGUISTIC EVOLUTION
LINGUISTIC EVOLUTION 147
cases proliferation of some consonant clusters are
on record in the case of the so-called 'geminates', i.e. the
same consonant at the end and at the beginning of two
successive syllables.
In modern English, geminates do not exist except at the
juncture of two words, as in pen-knife /pen-naif/ or head-dress.
But in many languages, such as Italian, Hungarian, or
Finnish, geminates are perfectly normal within the same
moneme, Geminates do not seem to have existed in what is
reconstructed as Proto-Indo-European. But they were a
normal feature of ancient Greek, Latin, and the older Celtic
and Germanic languages. It seems that, at some time or other,
the speakers of every one of these developed a habit of replac-
ing single intervocalic consonants by geminates in words felt
to be 'expressive":' the word knock, for instance, attested in
Old English as cnocian with a single c, appears in Middle
English as knokke, with a geminate which accounts for the
quality of the modern vowel (from cnocian, one would have
expected *knoke rhyming with oak). In the course of centuries,
these geminates ceased to be optional, and, as shown by the
ME. knokke, became perfectly normal features of the words
where they appeared. When added to the bulk of the ones
that had developed through various phonologically regular
processes, their frequency became comparable to that of
their simple counterparts. In other words, the information
of intervocalic /-kk-/ became similar to that of intervocalic
/-k-j. In English, as in the other Germanic languages, the
burden of the distinction between /-kk-/ and /-k-/ was n r ~
ally shifted over to the preceding vowel: OE.. socian became
ME. soke, with the vowel lengthened, hence NE. soak with
[o"], while ME. knokke preserved its short vowel, hence the
non-diphthongized vowel of NE. knock. The distinction be-
tween j-kk-/ and /-k-/ could then disappear and the balance
between information and cost be restored. It is interesting
I Cf. A. Martinet, La Gemination consonantique d'origine expressive dans les langues
germaniques (Copenhagen, 1937), pp. 29-44,
10
4- 47.
to notice that at the
place in England, the as a nation had hardly any
contacts the bulk of users of other Germanic languages..
This indicates that the adoption of that particular solution
of the problem must have been by such general
features of the language pattern as the phonematic build of
words and the of vowels which at the
beginning of the second millennium after Christ must still
have been largely similar throughout the Germanic domain.
In Latin, the ofgeminates had begun earlier, I
at a time when word final vowels had not been reduced, as it
was in Germanic the process ofreadjustment started, and
this difference in the eCOl101l1Y of word forms con-
tributes to the divergent paths followed in this con-
nexion by Romance and Germanic. The assumption that
Gaulish, the linguistic substratum of most of Western
Romance, had a finger in the pie cannot be ruled out: the
main initial features of the consonantal evolution of Western
Romance and Brythonic Celtic are so much alike that one
is legitimately tempted to assume that Gaulish, another
Celtic form of speech, as it was being displaced by Latin, was
undergoing what has been called 'lenition', and lenition,
a weakening of intervocalic consonants, would be nothing
but the outcome of a situation where geminates had begun
to be practically as frequent as their simple intervocalic
counterparts. TIle Romance consonantal shift
would accordingly be nothing but a reflex of a Celtic shift.
Yet the structural conditioning of the Western Romance
change must have been largely present in late Latin, and
Celtic may been just the last straw.
The Western Romance outcome is a general simplifica-
tion of the geminates, as a rule without any confusion with
corresponding single consonants. In the case of stops and
I Cf. A. Graul', Les Consonnes geminees en latin (Paris, 1935), where the expan-
sion of gemination, supposed to be a Proto-lndo-European phenomenon, is
examined. On gemination in Romance and Celtic, see Economic, pp. 257-96.
LINGUISTIC EVOLUTION
the jg"ellllIlatie'I
tion: late Latin tottabecomes toute, whereas the single
consonant is shifted: Latin tota yielding Spanish toda. Nasals
and liquids, for which, in the absence of any opposition of
voice, the conditioning was largely different, present diver-
gent treatments which point to later evolution: Castilian has
palatal reflexes of /-11-/ and I-nn-j and retains 1-
1-1
and I-n-I
with their old values; Portuguese shifted the former to 1-
1-1
and I-n-I, while Latin single 1-1-1 and I-n-I were eventually
eliminated as segments, although the nasal tamber of I-n-I
and, possibly, the velar quality of 1-1-1, must have been car-
ried over to neighbouring vowels, as we see from siio-csanum.
French shifted the burden of the distinction on to the pre-
ceding vowel, whose quality differed according to whether
it was in a former open or checked syllable (OFr. pele 'shovel'
from Lat. pala, but balle from germanic balla, sainfrom sanum,
but pan from pannum). This is, of course, parallel to what we
have found for Germanic at a much later period. In the case
of I-rr-I versus I-r-I, the process of simplification is still at
work in many quarters. In the case of French, the burden
of the distinction could not, in this case, be shifted to the
preceding vowel because many /-rr-I were the reflexes of
former I-tr-I and I-dr-I, as in pierre, from petra, and lierre
l'ierre), from hed(e)ra. Since I-tr-I and I-dr-I did not 'make
position' in Latin, the preceding vowel was in an open
syllable and was treated as such, hence the diphthongs of
pierre and lierre. But since the vocalism preceding the gemin-
ated I-rr-/ in pierre was the same as that preceding the single
/-r-I ofjiere, fromftra, the burden of the distinction was left to
the consonant, It remained there, in some form or another, in
Parisian French, down to the seventeenth century, when
/-rr-I must have passed first to a long uvular r which affected
the tamber of some of the preceding vowels: today the inter-
vocalic consonant is the same in marron and parer, but most
older Parisian speakers have lal in the former, and I-a-I in
the latter.
LINGUISTIC E VOLUTION 149
soon as phonological changes are considered within the
frame of the language system, the whole of the consonantal
'-' ... \.AC"" ... '-J' ... .a. of Western appears as ultimately
conditioned by the economical necessity of reducing the
energy needed by the articulation of geminates to the amount
compatible with their informative value. It is not meant
hereby that other factors have not been at work. To say that
the difference between the Castilian and the Portuguese
treatment of /-11-/ is due to chance would amount to stating
that the factors that have determined its shift to /-1..-/ in one
language, /-1-/ in the other have not been identified yet. In
a number of cases where French deviates from its partners,
the more intimate contacts of that language with Germanic
forms of speech must have at work; lengthening and
diphthongization of accented vowels in open syllables seem
to be good illustrations of these, But the diverging treatments
are nothing but locally determined reactions to the unbalance
deriving from the high frequency and low informational value
ofgeminates, '
The existence of an inverse relationship between frequency
and linguistic complexity is abundantly illustrated on the
plane of grammar, I Statistically first, when it is found that
grammatical items, each of which is on the average incom-
parably more frequent than is the case with lexical items,
are as a rule much shorter than the latter. Minimal com-
plexity, for a moneme, would seem to amount to formal
inexistence, and this indeed is what we find in the case of
such grammatical functions as are marked by position: in
languages with a subject, i.e, a formally characterized com-
pulsory actualizer of the predicate, no normal complete
statement can exist without one. This determines a maximal
probability which reduces the information to zero. Of course
the choice of a given subject among all the ones available and
I Cf. A. Martinet, 'Linguistique structurale et grammaire comparee'
Traoaux de l'Institut de linguistique, i (1956), pp. 7-21. '
LINGUISTIC EVOLUTION
likely in the particular context may furnish a large amount
of information. But the presence of a subject as such affords
none whatsoever, and it is normal that it not be posi-
tively marked by a specific speech segment or any formal
modification. Consequently, a zero ending for 'subject-case'
is what should be expected. We find a similar situation when
we consider the 'object'. The object is not 'quite as compul-
sory as the subject: eat is never without a subject, except ill
the imperative, but it may appear without an object. Yet,
once a given 'transitive' verb is chosen by the speaker, the
only information furnished by the object as such results
from the fact that it might have been suppressed. But in
contexts where an object is actually found in 99 per cent. of
the cases, its information is close to nil if considered in its
object-function, and not as a specific choice among a variety
of lexical or pronominal items. It is, therefore, normal that
the most frequent type of object should not be marked by any
specific segment, provided it is kept distinct from the subject
by means of word order.
This indeed corresponds to what we find in a majority
of languages with subjects. But there is an exception that
certainly looms large for classical scholars, namely the posi-
tively marked nominatives and accusatives of classical Indo-
European languages. If we leave out the plural cases which
raise special problems, we may state that, with the exception
ofsome neuters, the accusative, which is the case ofa majority
ofverbal objects, ends in -me The nominative, often identified
as the subject-case, ends in -s in the majority of masculine
and feminine nouns. Ifwe oppose masculine and feminine to
neuter as animate opposed to inanimate, we may say that
animate nouns ending in stops and prehistoric vowels present
an -s nominative (Latin pleb-s, cioi-s, ficu-s); others, which
have what is often called a zero ending, have a stem-final
consonant or an -a which must be the reflex ofvowel +'laryn-
geal' , the latter probably some fricative continuant. This
points to a prehistoric stage were <S was preserved after stop
LINGUISTIC EVOLUTION
or vowel, but dropped after a continuant (hence Greek
animate nouns with final -Yjp, -wv, -Yj -0: < *... eH)), and
a previous stage when ... s was the universal ending of the
nominative of animate nouns.
Whether the preceding reconstruction is accepted or not,
we are faced with the existence, in early Indo-European
languages, of an -s nominative limited to the designation of
beings or objects conceived as animate. This limitation sug-
gests that in Proto-Indo-European this -s marked a function
felt to be characteristic of animate creatures, namely that of
acting, as opposed to the passivity of things. Such -s forms
could hardly be subjects, i.e, forms used with the function
of compulsory actualizer of a predicate, because the zero
information of a subject as such would hardly be compatible
with a formally existing indicator. All this points to -s as the
mark of an ergative case, the case of the agent, which must
have coexisted with a true nominative, the form of the noun
used in order to introduce or 'nominate' a person or to pre-
sent a creature or an object. This true nominative, being used
outside of grammatically organized utterances, was in no
definite relation to a predicate and had therefore no gram-
matical function properly so called. It must have been
identical with the bare stemof the word and, as such, identical
with the vocative which also corresponded to a use of the
noun outside of grammatically organized utterances. In
languages that make use of an ergative case for the agent, I it
is frequent, ifnot universal, to find the bare stem used for the
most obvious complement of the predicate; in an utterance
corresponding to 'the woman is washing the linen', the
woman, as the agent, will be a complement in the ergative,
but the linen, as the 'most obvious complement', will be in
the nominative. The nominative, here, is no longer the sub-
ject case, but the bare stem as used in non-grammatical
I Basque is an example; for an evaluation of the ergative construction in
Basque, see the author's paper 'La Construction ergative et les structures
elementaires de l' enonce', Journal de psychologie normale et pathologique (1958),
PP377-92
LINGUISTIC EVO N LINGUISTIC E VOLU'TION
contexts: in a word-for-word translation, the equivalent of
the former utterance would sound like washing [if] (the)
linen by (the) woman or linen washing by woman. intran-
sitive predicates, 'the most obvious complement' is identical
with what we call the subject: 'the man talks' is presented as
talking [if] (the) manor mantalking, where manis treated as the
normally expected complement, and consequently presented
without explicit mark of its function.
A similar situation must have existed at some stage of
Proto-Indo-European: the bare stem was used out of context
as a nominative properly so called and as a vocative and, with
a grammatical function, as the most obvious complement
corresponding to the direct object of our transitive and the
subject of our intransitive verbs; the -s form was used for
marking the agent, but there could be an agent only if some
being or object (explicit or understood) was being acted upon:
in equivalent of man talks', the man was not treated
as an agent. The working of this is clarified if we think in
terms of nominal predicates and parallel talking if man and
washing oflinen. Nouns designating objects or beings that were
never conceived of as agents would not get a chance to
receive the -s tag. We are so used to identifying agent and
subject, that it is difficult for us to remember that an instru-
ment as such can never be an agent. But the speakers of
Proto-Indo-European would not have been misled into using
the designation of the sieve (Lat. cribrum, GE. hrider n.) in the
ergative -s case; a sieve in action would, of course, be used in
an instrumental case. IE. neuters are properly the nouns
which, on account of their meanings, were never used in the
ergative. some lin.guists would put it, the absence of an -s
form must have been unfait de parole before it became unfait
delangue. A neuter noun like Latin mare is used as a bare stem
both in mare videt 'he sees the sea' and mare patet 'the sea is to
be seen', and this mirrors a former stage of Indo-European
syntax.
shift from a former ergative construction to the linguis-
tic form with a compulsory subject, which we find in attested
languages, may have been determined by the develop...
ment of the a subject we cannot investigate here.
The use of the rS form instead of the bare stem may have
expanded first to constructions with intransitive verbs and
ultimately to the properly nominative uses. The gradual
elimination of the bare stem from its vocative uses and its
replacement by the -s form is an historical process which is not
quite completed yet: in Czech, for instance, the vocative
clovece 'man!' is still distinct from the nominative c!ovek (whose
-s, by an irony of fate, has long since disappeared). Formally
this expansion of the -s form resulted in a twofold anomaly:
the <S form was by now a subject, i.e. a form with no gram-
matical information any more, but with a functional mark still,
and the bare stem of animate nouns, restricted now, in gram-
matical contexts, to the function of complement of transitive
verbs, was deprived of any grammatical sign although its
occurrence was somewhat less automatic than that of the
formally marked subject. This latter aspect of the anomaly
was soon remedied through the replacement of the bare-
stem by a to-case in -m, the attested accusative (cf. Spanish
'direct' animate objects in a: ueo a Pablo '1 see Paul'). This
was analogically extended later to neuter o-stems.
The other lack of balance between information and cost
was more difficult to eliminate: -s forms, the bulk of the new
nominatives, were extremely frequent, early learned, and
consequently resistant: if word final s dropped out at all, it
could only be as the result of a regular phonological change,
but not just in the particular cases where it was the nomina-
tive function marker. Its regular prehistoric elimination after
sonants was welcome: it may well have been dropped only
in certain contexts (utterance finally, for instance, or before
consonant ofa following word), but, ifso, the forms without
-s must have been eagerly extended by analogy to all posi-
tions. In thematic nouns, i.e, after the elo vowel, it is well
attested in many Indo-European branches. But, in later
154 LINGUISTIC EVOLUTION
LINGUISTIC EVOLUTION
stages, whenever speakers are given a chance to choose,
among phonologically regular forms between an -s form and
an -r-less form, they give preference to the latter. Still, in
many cases, centuries elapsed before phonological evolution
offered any such chance: in French, one had to await the
elimination of declension before -s singular nominatives
yielded to -s-less oblique case forms.
In the lexical domain, abbreviation of segments whose fre-
quency is on the increase does not raise problems, as a rule,
if those segments are phrases of the underground railway type
where the more general term can be left out and the specifier
preserved as the equivalent of the whole. The reverse may
also happen, the more general term being preserved, if the
object that the phrase designates becomes so universal that
the general term may be used, without any danger of con-
fusion, as its normal designation: in most countries, motor-
cars were first referred to as 'automobile carriages'; this
phrase was soon curtailed to automobile and, finally to auto.
But nowadays, after the practical elimination of all horse-
drawn carriages, the usual designation of motor-cars is the
old word for 'carriage'; Fr. voiture, Germ. Wagen, and so
forth. If the designation is a learned word made up of
elements that were significant in Latin or Greek, but make
little sense for the average user of modern languages, the
curtailing will be made at random, as far as etymology is con-
cerned, but in such a way as to get a handy form whose size is
well suited to its frequency and which rules out any danger
of homonymic confusion..
Another frequent trick is the use of initials, as in LCC or
USSR. These are particularly favoured among specialists,
but are often resented, not only by purists, but by large sec-
tors of the general public. Still they often represent the
normal way out of some informational quandary: when the
London County Council became a generally identified con-
cept that had to be frequently referred to by many people,
its frequency in speech became such that the cost of its full
designation exceeded by far what corresponded to its, by then,
informational No curtailing was possible
since the combination of any two of its elements would have
been misleading. The use of the three initials, probably first
in writing, was certainly the most obvious and, in this com....
plex world of ours, the most natural solution.
In common speech, where no spelling can suggest the use
of initials, there may occur similar situations where none
of the components can be left out, whatever the increase in
frequency. Since the phonematic bulk cannot be reduced,
speakers resort to a reduction of the morphological com-
plexity and coagulate some of the elements into a single
maneme or, at least, handle these elements grammatically as
if they were one. The French phrase avoir l' air 'to look (like)'
affords a nice illustration of coagulation in two successive
stages. This phrase, in daily speech, must be about as frequent
as its English equivalent and it has practically eliminated its
traditional and literary competitors sembler and paraitre. In
its most usual form, the third person singular of the present
il a l' air jilaler j, it cannot be said to be phonologically more
complex than what its frequency would lead us to expect,
but its semantic articulation is cumbersome, as is shown by
the behaviour of speakers: in 'unguarded' colloquial French,
it is made to behave as a sort of copula; for 'she looks nice',
one would expect elle a l' air gentil, with gentil agreeing with
the masculine air; what one actually hears is elle a l'air gen-
tille, with gentille agreeing in gender with the subject elle.
With many children, the process of coagulation goes further:
air is no longer identified as a noun, but as the second syllable
of the stem alair-jaler-j of a regular -erverb *alairer, hence, in
the imperfect, fa m'alairait bon, 'it looked good to me', instead
ofstandard fa m'avait l' air bon.
We have, so far, been considering what happens to a form
when its frequency increases. In cases where it decreases, we
have pointed out that this does not necessarily involve any
LINGUISTIC E VOLUTION
LINGUISTIC EVOLUTION 157
formal change provided no homonymic conflict arises .. But a
lowering ofthe frequency which, in itself, would not endanger
the survival of a form, may seal its doom if it is somehow
isolated in the structure, if, for instance, it is a verb that is
not inflected according to a widespread analogical pattern.
When children learn their language, they tend to imitate
comparatively long stretches they have heard before they are
able to analyse them, i.e .. to use in other contexts the elements
they contain. As long as they do not use analogy, which
amounts to contriving new moneme combinations, they have
no problems with irregular forms .. But when they start doing
it extensively, which, with the average child, will begin about
the age of five, they will be apt to replace an irregular and
isolated brought by an analogical brung, Irregular forms that
are. very frequent are, by that time, well entrenched, and
occasional slips like brung will eventually disappear. But
those whose frequency is somewhat lower will be widely
mishandled. If the community is not of a conservative type,
children will be allowed to use their own forms, and ana-
logical forms will finally get established. on the contrary,
adults insist on correcting children, the latter will try to
avoid the forms they do not know how to inflect to the satis-
faction of their elders; these trouble-makers will become
obsolescent and eventually disappear. English move was bor-
rowed from French and was from the start inflected accord-
ing to most general analogical patterns. It is today one of
the most frequently used lexical items of the English language.
The original French mouvoir must have had, from early times,
a number of unusual inflexional features. In any case, there
is today no other verb conjugated just like it. It is about as
peculiar in its inflexional comportment as pouvoir or couloir,
yet it must always have been somewhat less frequent than
these quasi-auxiliaries, and the amount of irregularity which
speakers could cope with in handling these proved a little
too high for mouvoir. As a result, mouvoir has disappeared from
colloquial French; the normal equivalents of 'move' are
remuer, bouger, demenager, all single-stem -er verbs; another
literary equivalent of move is emouooir which survives, in
speech, as an two participles: emauuant and'
emu; but even these are threatened by the forms of the
denominative single-stem emotionner. I
We have so far, in the case of significant units, been
operating with general frequency. But in order to under-
stand how language works and changes, it is quite essential
to operate with frequency or, better, probability in given
contexts: book and walk are frequent words, but a combina-
tion like the book walks is quite unexpected and either non-
sensical or highly informative because it is quite improbable.
In other words, the informative value of the book walks is
higher than what a calculation based upon the general fre-
quency of book and that of walk might lead us to expect. On
the other hand, blue and sky are so frequently used together
that blue sky contributes less information than the average
probability of blueand that of sky would suggest. According to
whether the speaker or the writer wants to startle his public
or, on the contrary, to flatter his laziness and conservatism,
he will try to find unexpected combinations, or use time-
hallowed phrases and turns. Generally speaking, cultured
audiences or publics are those who, at the same time, are able
to digest more information per second, and need more un-
expected combinations in order to get the same amount of
information, since they hear new turns more often and con-
sequently find them much sooner hackneyed and stale. It is
thus clear that style is largely a matter of density of informa-
tion, and this covers to a large extent the subject-matter ofthe
experience, real or fictive, which is communicated, since the
understanding and appreciation of literary communication
implies previously shared experience, and shared experience
implies familiarity with a certain vocabulary. Whether com-
munication is of a literary nature or just an everyday spoken
I For an illustration of the working of formal analogy, see 'De I'economie
des formes du verbe en francais parle', Studia ... in hone L. Spitzer, pp. 3I 5 ff.
LINGUISTIC EVOLUTION
LINGUISTIC EVOLUTION 159
affair, it involves a constant renewal of the details of lin-
guistic practice, because everyone of us, author or speaker,
will find it necessary to emphasize some segments of his
utterances through additional information resulting from
unexpected collocations. If the trick proves successful, the
inventor himself or some of his hearers will be tempted to
use it again, and the new turn may get established in the
language. But should it become too popular, it would soon
lose all its efficiency and be replaced by some fresh innova-
tion. There are, at all times, large sections of the vocabulary
that are in a constant state of unrest, and since all the parts
of a language condition one another, we find here a per-
manent cause of linguistic evolution.
It is tempting to try to reduce all linguistic facts to quanti-
tative data, as we have been doing. But we should not forget
that what we may call the nature or quality of the various
linguistic units plays an essential role in the conditioning of
linguistic evolution: what can be measured is the quantity,
not the quality of information. If a new word enters a
language, its appearance will necessarily modify the in-
formational economy of that language. If there were n words
before, there will now be n+I words. In theory, everyone
of the n words should have become less frequent in the pro-
cess. But, in fact, the frequency of all but very few of the
n words will not be affected. Only those whose meaning is in
some way connected to that of the new-comer are likely to
score a lower probability. The impact of the lexical expan-
sion will be limited to a certain semantic domain. Besides, we
cannot be satisfied with the statement that the appearance of
a new word A has reduced by so much the probability of a
formerly existing word B, because this does not tell us what
semantic field remains allotted to it and what section of its
former domain has been encroached upon by A. This we
need to know if we want to understand the further repercus-
sions of the adoption of A. In the field of phonology, we have
seen that a similar frequency of geminates and corresponding
single consonants is only one of the factors that determined
the simplification of the geminates: simplification can
only occur if no dangerous confusion will ensue, which
largely depends on the phonic nature of the other units of the
system. All this means that applying the statistical technique
of informational research to language should not make us
forget that semantic and phonic properties of linguistic
units cannot be disregarded when problems of evolution are
at issue. The importance of the implications for our research
of recently evolved methods of investigation should not entail
a disregard of previous efforts towards the understanding of
semantic evolution. I
As regards phonology, whole-hearted attempts at estab-
lishing a structural method of diachronic investigation are
comparatively recent, and they have, if not from the start, at
least very early, involved informational considerations." A
basic problem, with phonological changes, is to explain how
the changing needs of man can affect the second articulation
of language, one of the main functions of which is precisely
to make the phonic form independent of the semantic value
of the message and of its significant components. The hypo-
thesis that frequency could modify cost, i.e. exert an influence
on the form of the units, phonemes as well as monemes,
suggested that this was at least one of the channels through
which communicative needs could act upon the economy of
phonematic patterns. Yet there are other channels, and we
must keep in mind the role played by prosodical features
such as accent and intonation which, being direct responses
to communicative needs, are most likely links between these
and the phonemes." But what is probably more important
than determining how external factors of unbalance pene-
I Cf. Suzanne Ohman's survey paper 'Theories of the "Linguistic Field" "
Word, ix (1953), pp. 123-34.
2 See Economic, pp. 139-47.
3 cr. Manual of Phonetics, ed. by Louise Kaiser (Amsterdam, 1957), pp. 255-
62.
160 LINGUISTIC EVOLUTION
trate the system is the examination of the repercussions of
that unbalance within the system. When trying to account
for past linguistic changes, on anyone of the planes of
language, there will come a point where we will be at a loss
to go farther for lack ofprecise historical information. It is not
suggested here that we should not try to cull useful hints
wherever we can find them, but we shall achieve the best
results if we concentrate on a domain we know well and are
best equipped to investigate, namely language structure In
all its aspects.
I E
accent, 34-35.
'accent', viii, 38.
accentual typology, 85-87.
accidence, go.
accusative (IE.), 153.
active, cf. voice (of verbs) .
actualization, actualizing, actualizer,
42 , 48, 61-63.
adjective, 64-65.
amalgam, amalgamation, 46, 47, gl.
analogy, 156.
arbitrariness, 88.
autonomous moneme, 45.
autonomous phrase, 44.
autonomy (syntactic), 44-45, 52.
bilingualism, 108, 113.
bilingualism (dialectal), 116-20.
case-ending, 45-46.
centrifugal, centripetal, 52.
change (causality of), 136-44.
choice,8-g, 16, 17, 30.
clause (main, subordinate), 50.
coagulation, 155.
coalescence (syntagmatic), 93.
colloquial speech, 124.
communicative function (of lan-
guage), 21.
community (language), 103-8.
complement, 49.
composition, 94-95.
concord, 53-55.
conjunction, 45.
content, 39-40.
contours (intonational), 30-32, 35.
convergence, 72, 104
core system (of phonemes), 78.
correcting children, 156.
cost, 136.
dependent (moneme), 45.
derivation, 94-95.
determinants, 50.
'determination', 64.
dialect, I I I
811928
s
dialectj, 112-18.
dialects, 112-18.
discontinuous signifiant, 92.
discrete units, 24, 33, 37, 39
distinctive features, 30.
double articulation, 24, 26, 28, 30, 59.
duration, 3I .
economy (linguistic), 2, 139, 141.
ergative, 151.
expansion (paradigmatic), 95.
experience, 21.
expression, 39-40.
expressive function (of language), 2I
first articulation, 22-23.
France (socio-linguistic situation in),
cf. socio-linguistic.
frequency, 51, 135
frequency/information, 143.
function, vii-viii, 3.
function of monemes, 49.
functional indication, 97-98.
functional moneme or indicator, 45.
functionals v. modifiers, 95-97.
geminates, 146, 154-7.
gender, 15-19.
gestures, 28, 30.
glossematics, 39-40.
glottal stop as a tone, 3I, 85.
glottalized consonants, 32.
glottis, cf. vocal chords.
grammar, 100.
grammatical monemes, 50-51.
grammatical typology, 89.
grammaticality, 97-98.
hierarchy among linguistic opposi-
tions, 107.
hierarchy (functional), 29.
idiolect, 105-6.
'idioms', viii.
inertia, 138, 139.
information, 46, 142-3.
M
162 INDEX OF SUBJECTS
INDEX OF CTS
106-7
vocal nature
voice, 31.
voice (of verbs), 49, 51, IOI.
word, go, g2.
word order (relevancy of), 4 I, 64.
writing, 25-26.
written v. spoken language, 122.
0.1"-"'.""1"<,<-,,,,,,9 1- 93.
one community,
variadon (as a characteris tic of Ian-
universals (relational), 58.
quadrangular vowel patterns, 79.
quality v. quantity, 158-9.
phoneme, 6, 23-25, 27--28, 30 , 35-
36, 39-41
phonetic change (regularity of), 25.
phonology, 30.
pitch, cf. melody (speech).
position, cf. word order.
predicate, predicative, predicative
function, predicative moneme, pre-
dicative phrase, 44, 48-49, 62-63.
'predicatoid', 50.
preposition, 45-46.
primary dependents, 50.
primary function, 50.
probability in context, 157.
prosodic, prosody, 29-32.
prosodical typology, 84-87.
redundancy, 56,68,140.
language, 20, 26, 70.
least effort, 55, 139.
lexical monemes, 50-51.
lexical typology, 87-89.
liaisons, cf. linking.
linearity of speech, 21, 22, 23, 24, 25.
linguistic communication, 21-22.
'linguistic field', 159.
linguistic reality, 70.
linking, 127-8.
literary communication, 157-8.
literary language, 105, 124.
informationjcost, 143-5.
information theory, 135,
inseparability, 92-93.
intonation, 35-38.
intonational habits, 37-38.
intonational levels, 35-37.
inventory (moneme), 50-51.
isomorphism, 39.
main clause, cf. clause.
marginal dependents, 50.
margins, 19-20, 28.
meaning, 95-96.
melody (speech), 27-35.
minimal utterance, 58-61.
modifiers, 51-52.
Inoneme, 22, 24,27-28, 30, 36, 39-41.
'morpheme', 22.
morphology, 89-90.
motivation, 88.
'mute e' (in French), I I, IS, 20.
neuter (IE.), 152.
neutralization (of phonological oppo-
sitions),8.
nominative, 151.
non-primary function, 50..
number, 17.
object, 44, 150.
operational devices, 5.
orders (of phonemes), 77,81-82.
out-of-situation, 60-62.
passive, cf. voice (of verbs).
patois, I I I, 117.
phonematic(s), 29-30.
phonematic typology, 74-84.
second articulation, 23.
series (of phonemes), 77, 81-82.
'shifters', 124.
signifiant, 22, 24, 30.
signifi, 22, 30.
signs (linguistic), 20, 22, 30.
simplicity (principle of), 9-10, 14.
situation, 59, 123-24
socio-linguistic situation in France,
113
solidarity between phonology and
grammar, 68-69.
spelling (English and French), I2S-6.
standard language, 112.
stress, 3I.
'stress', 35.
structure, 3-5.
subject,44,61,62-63,g6,IoI,I49-S0.
subject-predicate phrase, 48-49, 58.
subordinate clause, cf. clause.
successivity of linguistic elements, cf.
linearity of speech.
suprasegmental, cf. prosodic.
syllabic nucleus, 30-3I.
syntactic autonomy, cf. autonomy
(syntactic) .
syntax Chap. II, 8g-90.
tonal typology, 84-8S-
'tonemes', cf. tones.
s s
Abaza, 74.
Adyghe, 79.
African (Central . . . languages), 85.
Arabic, 78, 79, 85, 122.
Basque, 98, 119, 122.
Breton, I 19.
Brythonic Celtic, 147.
Celtic, 146.
Chinese, 85.
Creoles, 120-I.
Czech, 47, 86.
Danish, 47, 77, 79, 80, 85
Dutch, 29.
English, 6,7,8, II, 16, 17, 18,22,33,
4
2-43,44,45,63,73,75,77,78,79,
80,81,82,85,86,87,89,91,99,101,
107, 125, 156.
Ewe, 121.
Finnish, 78, 79, 146.
Flemish, 120.
Franco-Provencal, 109-10, 11'6-18.
French, 9-19, 22, 29, 33, 45, 58, 74,
75,76,77,78,79,80,81,82,84,87,
88-89, 91, 92, 94-95, 99, 104, 106,
107, 108-9, I I I, 125, 126-33,
148-9, 154, 155, 156-7.
Gaulish, 147.
Georgian, 75.
German, 8, 22, 23, 29, 45, 75, 78, 79,
80, 83, 86, 87, 88-89, 92, 120, 125,
154
Germanic, 146-7, 148-9.
Ghiliak, 33.
Greek (classical), 81, 82, 83, 146, 151.
Hawaiian, cf. Polynesian.
Hebrew, 76.
Ibo, 85.
Indo-European (Proto-), 146, 151,
152.
Indo-European languages, 150.
Iroquois, 82.
Italian, 78, 79, 80, 81, 87, 91, 146.
Latin, 44, 86, 87, 91, 92, 122, 138,
146,150
Latvian, 86.
Lithuanian, 85, 86.
Malagasy, 63.
Middle English, 146.
Norwegian, 30.
Old Church Slavic, 75.
Old English, 146.
Picard, 115-16.
Polish, 86.
Polynesian, 74, 82.
Portuguese, 146.
Romance, 146.
Romanian, 79, 80.
Russenorsk, 105.
Russian, 8, 74, 78, 80, 85, 86, 87, 100.
Sanskrit, 122.
Serbo-Croatian, 84, 85.
Spanish, 6, 74, 77, 78, 79, se, 81, 86,
14, 148.
Swedish, 30, 85, 86.
Tahitian, cf. Polynesian.
Turkish, 79.
Vietnamese, 64, 85.
Welsh, 23.
Western-Romance, 147-9.
'Volof, 121.
Zulu, 71.
Allen, vv. S., 4, 74.
Charles, 88.
Charles, 67, 73.
.IlJV.'JliJlF''-'i, Dwight L., 35.
1
5-
Cohen, Marcel, 66.
Dorfman, Eugene, 78.
Edmont, E., 114.
Fant, C., 75.
Gillieron, J., 114
Oougenheim, Georges, 52.
Grammont, Maurice, 12.
Graul', .A., 147.
Greenberg, Joseph, 67.
Han, Robert A., Jr., 121.
Halle, Morris, 75.
Haudricourt, Andre G., 74.
Hjelmslev, Louis, 2.
Jakobson, Roman, 10, 75.
Jespersen, Otto, 53-54, 124.
Jones, Daniel, 35.
Kaiser, Louise, 159-
Kurylowicz, ]erzy, 39.
Loftman, Beryl, 121.
____,... ' ,10.
Andre, 2, 9, I I, 14, 30, 72,
80, 84, 106, 117, 126, 129, 146,
149, 151, 157
Antoine, 66.
Suzanne, 159.
Pittman, Richard S., 52.
Sever, 114.
.lL'- ......1\..-.1.I.:llA..1.1.1, Ruth, 9.
Sapir, Edward, 49, 56, 58, 66-67, 70,
93- 10 0
Saussure, Ferdinand de, I, 2, 22, 88.
J.,II4
Suzanne, 121.
Douglas M., 121.
Togeby, Knud, 14.
Trubetzkoy, Nikolai S., 74, 78-79-
Voegelin, Carl, 78.
Vogt, Hans, 76.
Weinreich, Uriel, 121.
Whitfield, Francis, 2.
Yegerlehnerv ]., 78.
George Kingsley, 2, 139, 145.